Web Developers
15-1254.00Median wage $92,650/yr70,190 employed (US)Rank #39 of 923 scored · top 4% by substitution
Develop and implement websites, web applications, application databases, and interactive web interfaces. Evaluate code to ensure that it is properly structured, meets industry standards, and is compatible with browsers and devices. Optimize website performance, scalability, and server-side code and processes. May develop website infrastructure and integrate websites with other computer applications.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution60
Exposure52
Augmentation81
Substitution — the headline: capability discounted by cost, barriers and adoption.
Exposure — technical capability alone, regardless of whether anyone deploys it.
Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.
Tasks on the substitution scale
29 rated tasks, binned by substitution score.
Position among all scored occupations
Distribution of 923 occupation scores; the marker is this occupation.
Tasks with substitution ≥ 70
41%
Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.
Why this score
The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.
Task automatabilityw 35%52
panel mean rating 3.1/5 → substitution pressure 52/100
Technical feasibility todayw 20%52
panel mean rating 3.1/5 → substitution pressure 52/100
Cost vs. human wagew 15%63
panel mean rating 3.5/5 → substitution pressure 63/100
Adoption barriersw 20%inverted — strong barriers lower the score75
panel mean rating 2.0/5 (barrier strength) → substitution pressure 75/100
Sector adoption velocityw 10%69
panel mean rating 3.8/5 → substitution pressure 69/100
Task breakdown (29 tasks)
Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.
Renew domain name registrations.
96CI 92–100 · exposure 100 · augmentation 38 · importance 3.7/5 · click for rater detail
Renew domain name registrations.
96| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Web development and IT operations teams already use automation extensively for routine infrastructure tasks. Domain renewal automation is well-established practice in software development contexts and managed hosting platforms, showing high adoption velocity in digitized organizations. |
| Sector adoption velocity | claude-sonnet-5 | 5/5 | Auto-renewal is already the default, near-universal practice across web hosting and domain management, representing essentially complete adoption of automation for this task. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by flagging approaching renewal dates, preparing renewal checklists, or generating notifications, but the core task (authentication and payment execution) is better suited to full automation than human-assisted workflows. Augmentation value is moderate since the task itself is so administrative. |
| Augmentation potential | claude-sonnet-5 | 2/5 | Since the task is already essentially fully automated as a background system function, there's little remaining role for AI to 'augment' a human performing it manually. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Domain renewal is a straightforward, highly structured task with clear inputs (domain name, registrar credentials, payment details) and outputs (renewal confirmation). Current AI agents can reliably retrieve renewal dates, authenticate with registrar APIs, process payments, and document confirmations—easily achieving 50%+ time savings with minimal setup. |
| Task automatability | claude-sonnet-5 | 5/5 | Domain renewal is a highly routine, rules-based administrative process (checking expiration dates, submitting payment, confirming registration) that is easily scripted or automated via registrar APIs and auto-renewal settings. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While renewals require valid payment credentials and registrar account access, there are minimal legal or licensing barriers. The main friction is credential management and organizational policy oversight rather than regulatory prohibition on automation. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, liability, or human-judgment requirement exists; domain renewal is a pure administrative/payment transaction with no regulatory or professional gatekeeping. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI agent execution (API calls, minimal compute overhead) costs pennies per renewal, while human time spent on this administrative task typically costs $15–50 per domain. The cost differential is at least one order of magnitude in AI's favor. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated renewal via registrar systems costs essentially nothing beyond the registration fee itself, compared to a developer's time manually tracking and renewing domains. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature products and deployed automation workflows handle domain renewal at scale today. Services like IFTTT, Zapier, and direct registrar APIs with automation capabilities reliably execute this task in production environments with minimal error rates. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Every major domain registrar (GoDaddy, Namecheap, AWS Route 53, etc.) already offers automated auto-renewal and API-driven renewal workflows in production, used at massive scale today. |
Back up files from Web sites to local directories for instant recovery in case of problems.
96CI 91–100 · exposure 92 · augmentation 50 · importance 4.3/5 · click for rater detail
Back up files from Web sites to local directories for instant recovery in case of problems.
96| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Tech and web development sectors have deeply adopted automated backup and CI/CD practices; backup automation is standard DevOps practice, not experimental. |
| Sector adoption velocity | claude-sonnet-5 | 5/5 | Web development and hosting industries have near-universal automated backup adoption already; this has been standard practice for over a decade, not an emerging trend. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by recommending backup strategies, monitoring backup health, and alerting on failures, but since the core task is fully automatable, augmentation is limited to oversight and optimization. |
| Augmentation potential | claude-sonnet-5 | 3/5 | While the task is largely automatable, AI/tools can still assist developers in configuring backup strategies, monitoring failures, or troubleshooting restoration, though this isn't a task requiring ongoing human-in-the-loop judgment. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | File backup from websites to local directories is a straightforward, repetitive data transfer task that AI systems and scripts can fully automate end-to-end with scheduling, compression, and integrity checking, easily achieving >50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 4/5 | Backing up website files to local directories is a well-defined, repetitive procedural task that can be fully scripted or automated via existing tools (rsync, cron jobs, cloud backup services, CI/CD pipelines).This meets the time-saving bar easily with off-the-shelf automation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 1/5 | No licensing, regulatory, or liability barriers prevent automating file backups; this is routine infrastructure automation with no legal requirement for human involvement. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, regulatory, or human-judgment requirement is attached to file backup; it's a purely technical/administrative task with no legal sign-off needed. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated backup solutions (open-source tools, cloud services, scheduled agents) cost fractions of a penny per backup cycle, orders of magnitude cheaper than paying a developer's loaded wage to manually back up files. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated backup scripts/services cost fractions of a cent to a few dollars per month versus paying a developer's hourly wage to manually back up files, an order-of-magnitude or greater savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature backup automation products and tools (rsync, git, cloud-native backup services, cron/scheduled tasks, CI/CD pipelines) reliably perform this task in production at scale across organizations. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Mature backup products (cPanel backups, GitHub Actions, AWS S3 sync, plugins like UpdraftPlus, managed hosting auto-backups) reliably perform this in production at massive scale today. |
Respond to user email inquiries, or set up automated systems to send responses.
86CI 75–97 · exposure 87 · augmentation 75 · importance 3.7/5 · click for rater detail
Respond to user email inquiries, or set up automated systems to send responses.
86| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Web development and tech sectors are among the fastest adopters of automation and AI. Email automation, chatbots, and ticketing systems are in broad production use in this industry. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Web development and tech support functions are in fast-adopting digital sectors where automated/AI-driven customer communication tools are already widely deployed in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI drafting tools significantly assist developers in composing email responses, suggesting templates, and prioritizing inquiries. Developers remain in the loop to review and personalize, raising their handling speed and consistency substantially. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI drafting tools and smart-reply suggestions substantially speed up a developer's handling of email inquiries even when a human reviews or personalizes the final response. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | A large portion of routine user email inquiries can be handled by current AI systems using email classification, templated responses, and context-aware generation. However, complex or nuanced issues often require human judgment, preventing full end-to-end automation at the 50% time-saving bar for all inquiry types. |
| Task automatability | claude-sonnet-5 | 5/5 | Drafting email responses and setting up automated reply systems (autoresponders, chatbots, templated workflows) is already fully achievable with off-the-shelf AI and automation tools, easily saving over half the time at equal or better consistency. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers exist for automating routine email responses. The main friction is organizational (desire to maintain human touch, liability for incorrect automated responses) and customer expectation, rather than hard requirements for human involvement. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, liability, or regulatory requirement mandates a human respond to routine web development inquiries; organizations freely deploy automated responders. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference cost for email response generation is very low (pennies per inquiry), with modest integration overhead. This is substantially cheaper than paying a developer or support person to manually respond, likely a 5–10× cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated email response systems cost fractions of a cent per inquiry via API calls versus a developer's hourly wage, representing an order-of-magnitude or greater cost reduction. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (chatbots, email automation platforms, LLM-based ticket systems) reliably handle straightforward inquiries at scale. Production use is common in tech companies, though integration and error handling for edge cases remain material challenges. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Mature production tools (help-desk AI responders, email automation platforms like Zapier/Intercom/Gmail smart reply, LLM-based support bots) reliably handle this today at scale across many organizations. |
Write supporting code for Web applications or Web sites.
82CI 77–86 · exposure 75 · augmentation 100 · importance 4.6/5 · click for rater detail
Write supporting code for Web applications or Web sites.
82| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Web development is a high-digitization, information-sector occupation with rapid AI adoption; GitHub Copilot alone is embedded in millions of developer workflows. Actual displacement and productivity gains are documented and accelerating across software companies. |
| Sector adoption velocity | claude-sonnet-5 | 5/5 | Software development is among the fastest and deepest AI-adopting fields, with AI coding assistants integrated into mainstream IDEs and workflows at scale already. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI code assistants are transformative for supporting code tasks, dramatically reducing boilerplate writing time while keeping the developer in control of architecture and integration decisions. Developers report substantial productivity gains in real-world use. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI coding assistants substantially speed up writing supporting code while developers remain in the loop to review, test, and integrate, representing a clear productivity transformation. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI code generation systems (GitHub Copilot, Claude, ChatGPT) can automate substantial portions of supporting web code (utility functions, API integrations, boilerplate) with significant time savings. While complex architectural decisions or novel integrations may require human oversight, the majority of routine supporting code generation meets the ≥50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 4/5 | Writing supporting code (utility functions, glue code, boilerplate, API integration) is highly amenable to LLM code generation, with significant time savings for common patterns, though complex architecture and debugging still need human oversight. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Supporting web code authorship has minimal legal or regulatory barriers; developers alone decide whether to use AI assists. The only friction is organizational policy, code review requirements, and developer preference—none hard barriers to adoption. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There is no licensing, liability, or regulatory requirement mandating a human write supporting code; adoption is purely a matter of organizational choice and tooling. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI inference costs for code generation are negligible (cents per task), while a developer's loaded wage for equivalent output is $50–150+/hour. The cost ratio favors AI by one or more orders of magnitude when overhead and integration are factored fairly. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI coding assistants cost a small monthly subscription fee versus a developer's hourly loaded wage, delivering large multiples in cost savings per unit of code output, though human review time offsets some savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed products (Copilot, Claude with IDE integration, ChatGPT) reliably generate supporting web code in production environments. Error rates on straightforward supporting code are low enough for practical use; developers verify and integrate output at scale in real organizations. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed tools like GitHub Copilot, Cursor, and Claude Code are used in production by many developers to write supporting code reliably, though they still require review and occasional correction. |
Establish appropriate server directory trees.
81CI 79–84 · exposure 75 · augmentation 88 · importance 3.5/5 · click for rater detail
Establish appropriate server directory trees.
81| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Web development and DevOps teams show strong AI adoption for infrastructure and code generation tasks, with widespread use of Copilot, ChatGPT, and CI/CD automation. This aligns with fast adoption in the information/tech sector. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector for AI tooling, with scaffolding/boilerplate generation already widely integrated into IDEs and CLI workflows. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI significantly assists developers in this task by instantly generating scaffolding, suggesting best practices, and auto-completing configuration patterns, allowing humans to focus on validation, security, and organizational alignment rather than manual tree design. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly speeds up initial project setup and suggests best-practice structures, though developers still adjust for project-specific needs and existing conventions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can generate directory structures and scripts for server setup with high consistency, and tools like Terraform/Infrastructure-as-Code combined with LLMs can automate much of the work. However, task completion requires validation against organizational standards and context-specific requirements, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 4/5 | AI coding assistants and scaffolding tools can generate standard directory structures and configurations based on framework conventions with minimal human input, meeting the time-saving threshold for most standard projects. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Minimal regulatory or legal barriers exist for automating directory tree creation. Light organizational friction remains (code review, security sign-off) but these are process steps rather than hard blockers, and deployment is purely technical without licensing requirements. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, regulatory, or liability barriers exist for organizing server file structures; it's a purely technical convention-driven task. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI inference cost for generating directory trees and associated scripts is negligible (pennies per task) compared to a developer's loaded hourly wage ($75–150+), making AI-assisted or fully automated solutions orders of magnitude cheaper. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Generating a directory structure via AI/scaffolding tools takes seconds at negligible compute cost versus a developer's billable time to plan and set up manually. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (GitHub Copilot, Claude, specialized DevOps tools) reliably generate server directory structures and configuration code in production environments. Developers routinely use AI for scaffolding, though human review of security and organizational fit remains standard practice. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Tools like GitHub Copilot, CLI scaffolding generators, and framework boilerplate generators (e.g., create-react-app, Django admin) reliably produce directory trees in production today, though customization for unusual architectures still needs review. |
Document test plans, testing procedures, or test results.
77CI 70–84 · exposure 70 · augmentation 100 · importance 3.5/5 · click for rater detail
Document test plans, testing procedures, or test results.
77| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Software development and web development sectors show high AI adoption velocity; documentation generation is a standard use case in professional development teams. Copilot, ChatGPT, and Claude see widespread production use for documentation tasks in agile environments. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector for AI coding and documentation tools, with widespread integration into IDEs and CI/CD pipelines already occurring. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI significantly assists developers by auto-generating documentation templates, formatting test results, and producing comprehensive reports while developers review and refine content. This preserves human judgment over test significance while dramatically accelerating document creation. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools significantly speed up drafting, formatting, and summarizing test documentation while developers retain control over accuracy and final review. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can generate comprehensive test documentation from test scripts, logs, and procedures with high quality and significant time savings. LLMs excel at structuring, formatting, and synthesizing technical documentation, though some human review and domain-specific refinement remains typical. |
| Task automatability | claude-sonnet-5 | 4/5 | Generating structured test documentation from test cases, code, or execution logs is largely a text-synthesis task that current LLMs handle well, especially with access to test scripts or CI output as context. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No legal or licensing requirement mandates human-authored test documentation. Primary friction is organizational preference for human review and sign-off on test results that affect quality decisions, but automation of the writing task itself faces minimal structural barriers. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There are no licensing, legal, or human-sign-off requirements for internal test documentation, making this task easily substitutable. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Inference cost for generating documentation is negligible (cents per document), while a developer documenting test plans manually costs $50–150/hour. AI-generated documentation requires minimal oversight, achieving at least 10× cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Drafting documentation via AI tools costs a fraction of a developer's time compared to manually writing detailed test plans and reports, even accounting for review overhead. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed products (GitHub Copilot, Claude, ChatGPT) reliably generate test documentation templates, format results, and create comprehensive test reports in production settings. Tools integrate directly into development workflows with generally high quality output. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI coding assistants and QA tools can draft test plans and summarize results today, but they still require developer review and editing to ensure accuracy and completeness, so reliability in production varies. |
Provide clear, detailed descriptions of Web site specifications, such as product features, activities, software, communication protocols, programming languages, and operating systems software and hardware.
77CI 70–84 · exposure 70 · augmentation 100 · importance 3.2/5 · click for rater detail
Provide clear, detailed descriptions of Web site specifications, such as product features, activities, software, communication protocols, programming languages, and operating systems software and hardware.
77| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Tech and software development sectors are early adopters of AI tools for documentation and specification generation. Production use is already common in many development teams and enterprises. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a high-digitization, fast-adopting sector where AI-assisted documentation and coding tools have seen rapid uptake. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments developer productivity by auto-generating specification drafts that developers refine, review, and validate—a high-leverage productivity multiplier for knowledge work. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools substantially speed up drafting of technical specs, letting developers refine and validate rather than write from scratch. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can generate detailed technical specifications covering product features, protocols, languages, and hardware requirements with substantial time savings. Current systems handle these documentation tasks reliably, though human review for accuracy and completeness remains typical practice. |
| Task automatability | claude-sonnet-5 | 4/5 | Drafting technical specification documents from requirements or existing code is well within current LLM capabilities, especially with codebase access, though final accuracy checks still need human review. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing or legal requirement mandates human authorship of specifications. Light organizational friction exists (preference for human review, QA sign-off) but poses no hard barrier to automation. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, regulatory, or liability barriers prevent AI from drafting technical documentation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI inference cost for generating a detailed specification is orders of magnitude cheaper than paying a developer $75–150/hour for the same task, with minimal overhead. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating documentation via LLM inference is far cheaper than dedicating developer hours to manual spec writing, though some oversight cost remains. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed AI tools (Claude, GPT-4, specialized documentation generators) reliably produce technical specifications in production environments. Tools are widely adopted for generating specification drafts, though final validation by engineers is standard. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI coding assistants and documentation tools (e.g., Copilot, Cursor, ChatGPT) can generate spec drafts, but production use still requires human editing for accuracy and completeness; not fully autonomous in most orgs. |
Document technical factors such as server load, bandwidth, database performance, and browser and device types.
76CI 70–82 · exposure 70 · augmentation 100 · importance 3.4/5 · click for rater detail
Document technical factors such as server load, bandwidth, database performance, and browser and device types.
76| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Web development and DevOps teams in information and tech sectors have rapidly adopted automated monitoring, alerting, and AI-powered observability platforms; this is mainstream production practice. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software/tech sectors are fast adopters of AI tooling for documentation and DevOps observability, with many teams already using AI-assisted logging and reporting tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI systems dramatically augment developers' ability to navigate and synthesize performance data, enabling faster anomaly detection, report generation, and decision-making while developers remain the final arbiter of technical interpretation. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI significantly aids developers by summarizing logs, metrics, and performance data into readable documentation, greatly speeding up this routine but detail-heavy task. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can automatically collect, analyze, and document technical metrics from server logs, monitoring dashboards, and analytics platforms with minimal human intervention, easily meeting the 50% time-saving threshold for this data-driven documentation task. |
| Task automatability | claude-sonnet-5 | 4/5 | AI tools can pull metrics from monitoring systems and generate structured technical documentation with minimal human editing, meeting the time-saving threshold for most of the writing/synthesis work.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers exist; the main friction is organizational preference for human review of critical infrastructure documentation and integration with existing DevOps workflows. |
| Adoption barriers | claude-sonnet-5 | 1/5 | This is internal technical documentation with no licensing, regulatory, or human-contact requirements blocking automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Automated monitoring and AI-assisted analysis cost a fraction of manual documentation labor; continuous collection and synthesis are substantially cheaper than human review and write-up of the same metrics. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated log/metric summarization and doc generation is cheap relative to a developer's time spent manually compiling performance reports, though integration setup adds some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature monitoring and observability platforms (Datadog, New Relic, Prometheus) already integrate with AI-driven reporting and log analysis tools that reliably document these technical factors in production environments at scale. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like AI-assisted documentation generators and observability platforms with AI summarization exist and are used in production, but full automation of gathering and contextualizing all factors still requires human oversight and configuration. |
Evaluate code to ensure that it is valid, is properly structured, meets industry standards, and is compatible with browsers, devices, or operating systems.
74CI 61–86 · exposure 62 · augmentation 100 · importance 4.1/5 · click for rater detail
Evaluate code to ensure that it is valid, is properly structured, meets industry standards, and is compatible with browsers, devices, or operating systems.
74| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Web development is a highly digitized, fast-moving sector with deep and rapid adoption of CI/CD pipelines, automated testing, and linting tools. Nearly all professional web teams deploy such automation. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector with CI/CD pipelines integrating automated linting, validation, and AI-assisted code review as standard practice. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI code analysis tools significantly augment developer productivity by catching errors early, suggesting standards compliance improvements, and identifying compatibility issues before deployment, while developers retain control over decisions and architecture. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-powered linters, validators, and code review assistants substantially speed up developers' ability to catch errors and standards violations while they remain in control of final decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Static code analysis, linting, and format validation are highly automatable; however, evaluating architectural quality and compatibility edge cases still requires human judgment. Current AI systems can catch syntax errors, style violations, and many structural issues automatically, achieving substantial time savings. |
| Task automatability | claude-sonnet-5 | 3/5 | AI linters, validators, and LLM code review can catch syntax errors, structure issues, and common compatibility problems, but comprehensive cross-browser/device validation and industry-standard judgment often still require human review and testing setups. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or licensing barriers exist. Code review remains largely an organizational practice rather than a legal requirement, and most teams already use automated tooling. Human oversight is preferred but not mandated. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or legal requirement mandates human code review; adoption of automated validation tools is already widespread and unrestricted. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated linting and CI/CD code validation cost pennies per run versus human code review labor. Integration overhead is minimal since these tools are standard infrastructure in modern development pipelines. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated linting and AI code review tools run at a small fraction of the cost of manual code review time, though some human verification remains needed for edge cases. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature tools exist in production (ESLint, Prettier, SonarQube, Chromatic testing) that reliably perform syntax and standards checking at scale. AI-assisted code review systems are deployed, though they work best as augmentation; fully autonomous evaluation of nuanced compatibility issues remains less reliable. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed tools (ESLint, Stylelint, AI code review assistants like GitHub Copilot/CodeRabbit) reliably catch many issues in production, but full compatibility validation across devices/browsers still relies on dedicated testing tools and human oversight. |
Perform Web site tests according to planned schedules, or after any Web site or product revision.
72CI 57–86 · exposure 62 · augmentation 75 · importance 4.0/5 · click for rater detail
Perform Web site tests according to planned schedules, or after any Web site or product revision.
72| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Web development organizations have been aggressively adopting CI/CD pipelines with automated testing for over a decade; this is now industry standard in software development, especially in tech and digital-first companies. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software/web development is a fast-adopting sector for AI tools, with CI/CD pipelines increasingly integrating automated and AI-assisted testing as standard practice. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-powered test tools assist developers and QA engineers by intelligently generating test cases, detecting flaky tests, and providing detailed failure analysis, significantly boosting the speed and coverage of the testing phase while humans remain responsible for strategy and interpretation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially speeds up test case generation, bug detection, and regression testing, letting developers focus on interpreting results and handling complex scenarios. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems and testing frameworks can automate the majority of scheduled web testing tasks, including functional testing, cross-browser validation, performance checks, and regression testing. However, some edge cases and exploratory testing may require human judgment, preventing a full 5-rating. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can generate and run automated test scripts, check links, and flag visual regressions, but designing comprehensive test plans and interpreting edge cases still needs human oversight for full end-to-end coverage. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Adoption barriers are minimal: testing automation is a standard engineering practice with no licensing or regulatory requirements. The main friction is organizational inertia and the upfront effort to write automated tests, not legal or liability constraints. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or legal requirement mandates a human perform website testing; it's a purely technical task with no regulatory gatekeeping. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated testing is substantially cheaper than manual QA once initial test suite setup is complete; a single automated test run costs pennies and scales to thousands of test cases, versus hours of human labor per revision cycle. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted testing tools reduce manual QA time significantly, but licensing, integration, and human review of results keep costs roughly comparable to a lean human tester for nontrivial sites. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature testing automation tools (Selenium, Cypress, Playwright, TestNG) are widely deployed in production by organizations of all sizes. AI-powered testing platforms (Testim, Mabl, LambdaTest) now offer reliable end-to-end test execution and reporting, though occasional human review of failures remains common. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Tools like automated testing frameworks with AI-assisted test generation (e.g., visual regression tools, AI-augmented QA platforms) exist in production, but reliability varies and complex functional/UX testing still requires human validation. |
Install and configure hypertext transfer protocol (HTTP) servers and associated operating systems.
71CI 61–81 · exposure 62 · augmentation 75 · importance 2.9/5 · click for rater detail
Install and configure hypertext transfer protocol (HTTP) servers and associated operating systems.
71| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | DevOps, SRE, and cloud-native adoption in technology and financial sectors show rapid, deep penetration of infrastructure automation; major enterprises routinely deploy servers via CI/CD pipelines with minimal manual intervention. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | DevOps and cloud infrastructure automation (via IaC tools, CI/CD pipelines, and AI-assisted scripting) is widely adopted in tech/software sectors, which have high digitization and rapid AI tool adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-driven configuration management and infrastructure-as-code tools significantly amplify a developer's ability to provision and manage servers, allowing them to focus on policy, security, and exception-handling while automation handles routine deployment steps. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI coding assistants substantially speed up writing configuration files, diagnosing server issues, and generating deployment scripts, significantly boosting developer productivity while humans retain final control. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems and infrastructure-as-code tools can automate the majority of HTTP server installation and configuration (e.g., using Terraform, Ansible, Docker) with significant time savings. However, edge cases around custom configurations, security hardening, and troubleshooting may still require human oversight, preventing a full 5-rating. |
| Task automatability | claude-sonnet-5 | 3/5 | AI coding assistants and agents can generate installation scripts, config files (e.g., nginx/Apache configs), and troubleshoot common setup issues, but full end-to-end server provisioning still often requires human verification of security, network, and environment-specific details. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few hard regulatory or licensing barriers exist for automating server installation; organizational friction and desire for human validation of critical infrastructure are the main adoption brakes, but no legal requirement mandates human sign-off. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but security-sensitive infrastructure work often involves organizational approval processes and risk-averse IT policies that slow full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated deployment via existing tools (cloud services, open-source orchestration) costs orders of magnitude less than manual human configuration, requiring only minimal oversight and integration labor compared to the loaded wage of a web developer performing this repeatedly. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Once configured, AI-assisted scripting and automation tools (Ansible, Terraform, AI-generated configs) are far cheaper than manual sysadmin labor per server instance, though initial setup and validation still require some human cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (cloud CLIs, infrastructure-as-code platforms, containerization tools) perform routine server setup reliably in production at scale. Specialized deployment automation is well-established in DevOps, though complex or legacy system configurations may still require human intervention. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | DevOps automation tools and AI coding assistants (e.g., Copilot, infrastructure-as-code generators) are used in production for scripting server setup, but reliable unattended configuration of OS and HTTP servers at scale still needs human oversight for edge cases and security hardening. |
Design, build, or maintain Web sites, using authoring or scripting languages, content creation tools, management tools, and digital media.
71CI 61–80 · exposure 62 · augmentation 100 · importance 4.4/5 · click for rater detail
Design, build, or maintain Web sites, using authoring or scripting languages, content creation tools, management tools, and digital media.
71| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Web development sectors (SaaS, digital agencies, in-house tech teams) are among the fastest adopters of AI coding tools. Production use of Copilot, ChatGPT, and similar systems is now widespread in information and tech services, with high digitization and strong competitive pressure driving rapid adoption. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software/tech sectors are among the fastest adopters of AI coding tools, with widespread production use of AI-assisted development workflows already documented across the industry. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI provides transformative productivity assistance for web developers: code completion, bug detection, documentation generation, and rapid prototyping significantly accelerate development cycles while developers remain in control of architecture, design decisions, and quality assurance. This is one of the clearest cases of human-AI collaboration in augmentation. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI coding assistants substantially speed up scaffolding, debugging, styling, and content generation while developers retain control over architecture, integration, and final quality decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can automate significant portions of web development—code generation, boilerplate creation, debugging, and basic layout—but full end-to-end site design and maintenance still requires human judgment on architecture, UX decisions, and integration complexity. Roughly 40–50% time savings is achievable with current tools like GitHub Copilot and GPT-4, but reaching ≥50% consistently at equal quality requires structured setup and human review. |
| Task automatability | claude-sonnet-5 | 4/5 | AI coding assistants and agentic tools can now scaffold, code, and deploy full websites from natural language descriptions, handling much of the boilerplate and even complex logic with iterative prompting, meeting the 50% time-saving bar for many typical sites. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Web development lacks hard licensing or liability barriers; no legal requirement mandates human sign-off. However, organizational friction (QA processes, code review culture, client trust in human developers) and the need for human judgment on design and architecture decisions create modest adoption friction. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There is no licensing, regulatory, or liability requirement mandating a human web developer; adoption is purely a matter of organizational choice and quality control. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference costs for code generation and assistance are minimal (cents per task), while junior developer labor costs several hundred dollars per day. The all-in cost (API calls + human oversight + integration) remains substantially cheaper than hiring for routine coding work, though complex architectural tasks still benefit from human expertise. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI coding subscriptions cost tens of dollars monthly versus developer wages of $40-80+/hour, making AI dramatically cheaper per unit of output for routine web development tasks. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (GitHub Copilot, ChatGPT with code plugins, Vercel AI) demonstrably assist and partially automate web development tasks in production environments. However, error rates remain material for complex logic, and scope is narrower for full-stack design; most systems excel at code completion and refactoring rather than end-to-end site creation. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like GitHub Copilot, Cursor, v0, and various AI website builders are deployed and used in production, but they still require human review, debugging, and design judgment for anything beyond templated or simple sites. |
Monitor security system performance logs to identify problems and notify security specialists when problems occur.
69CI 57–81 · exposure 62 · augmentation 75 · importance 3.7/5 · click for rater detail
Monitor security system performance logs to identify problems and notify security specialists when problems occur.
69| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Security monitoring automation is deeply embedded in modern infrastructure: nearly all enterprises and DevOps teams use automated log aggregation and alerting as standard practice. Adoption is widespread across information technology, finance, and cloud-native organizations. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | IT and security operations are among the faster-adopting domains for AI-driven monitoring and alerting tools, with widespread production use of automated log analysis and SOC tooling. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-powered log monitoring significantly augments human security specialists by filtering noise, highlighting critical anomalies, and enabling faster response times. Specialists can focus on investigation and remediation rather than manual log review, raising overall team productivity substantially. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly enhances a developer's or security specialist's ability to sift through large log volumes, surface anomalies, and prioritize alerts, meaningfully boosting productivity while humans retain decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Log monitoring and anomaly detection are well-suited to automation: AI systems can parse structured logs, identify patterns deviating from baselines, and trigger alerts with high accuracy. However, the task requires some human judgment in contextualization and filtering false positives, preventing a full 5 rating. |
| Task automatability | claude-sonnet-5 | 3/5 | Log monitoring and anomaly detection can be substantially automated with SIEM/AI tools, but interpreting nuanced security events and deciding when to escalate still requires human judgment, so it meets partial but not full end-to-end automation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers prevent automation; security teams have strong financial incentives to adopt automated monitoring. The main friction is organizational preference for human oversight and integration complexity with legacy systems, not hard licensing or liability requirements. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No strict licensing requirement for this specific monitoring task, though organizational policies and liability concerns around security incidents create some caution before fully automating escalation decisions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated log monitoring is orders of magnitude cheaper than human security analysts reviewing logs 24/7. Cloud-based or self-hosted AI monitoring scales to handle millions of events per day at minimal marginal cost versus high-wage security specialist time. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Automated monitoring tools reduce headcount needs but still require licensing, integration, and human analysts for triage, keeping costs roughly comparable to a lean human-monitoring setup at moderate scale. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed security monitoring products (SIEM systems, log aggregators with ML-based anomaly detection) reliably perform this task in production at scale. Tools like Splunk, Datadog, and cloud-native solutions routinely detect and alert on security issues automatically, though integration complexity and tuning are required. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like SIEM platforms with AI-driven anomaly detection (Splunk, Datadog, Sentinel) are deployed widely, but false positive rates and need for tuning mean they are not fully reliable without human oversight. |
Develop system interaction or sequence diagrams.
69CI 59–80 · exposure 62 · augmentation 100 · importance 2.7/5 · click for rater detail
Develop system interaction or sequence diagrams.
69| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Software development and web development teams are actively adopting AI coding and design assistance, but diagram generation specifically remains a secondary use case with pilots outpacing full production rollout; adoption is faster in tech-forward firms than across the sector uniformly. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector for AI coding and design assistance tools, with many developers already using AI-integrated IDEs and diagram generators. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI substantially augments developer productivity by drafting initial diagrams, suggesting refinements, and accelerating iteration cycles while developers maintain full oversight and creativity over system design decisions. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI significantly speeds up the creation of initial diagrams and documentation drafts, letting developers focus on refining and validating logic rather than starting from scratch. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI can generate initial sequence and interaction diagrams from requirements or code analysis with substantial time savings, though oversight and refinement of complex or domain-specific logic patterns typically remain necessary, meeting the ≥50% time-saving threshold in many real-world workflows. |
| Task automatability | claude-sonnet-5 | 3/5 | LLMs can draft sequence/interaction diagrams (e.g., via Mermaid/PlantUML syntax) from requirements text, but accurately capturing correct system logic still requires human review and iteration, especially for complex or novel systems.deferring full autonomy. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing or legal requirement mandates human creation of diagrams; adoption is limited mainly by organizational adoption of AI tooling and developer preference for hands-on design in critical cases, not by hard regulatory or liability barriers. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There is no licensing, regulatory, or liability barrier to using AI-assisted diagramming tools in software development workflows. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | API costs for inference and storage of diagram generation are typically orders of magnitude cheaper than the fully-loaded cost of a developer's time to manually design and iterate on diagrams, especially for routine sequence flows. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating a first-pass diagram via AI is extremely cheap compared to a developer manually modeling interactions, though human validation still adds cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed products (Claude, ChatGPT, specialized diagramming tools with AI) reliably generate diagrams in formats like PlantUML, Mermaid, and UML pseudocode from natural language or code; production use is documented but usually requires human validation and iteration. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Tools like GitHub Copilot, ChatGPT, and diagram-generation plugins can produce draft UML/sequence diagrams today, but they are used as drafting aids rather than reliable standalone production systems for complex architectures. |
Perform or direct Web site updates.
69CI 57–80 · exposure 62 · augmentation 88 · importance 4.0/5 · click for rater detail
Perform or direct Web site updates.
69| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Tech-forward companies and web development firms are rapidly adopting CI/CD automation, code generation tools, and automated testing; production deployment automation is now commonplace in digitized sectors, though smaller or legacy-heavy shops lag. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Web development and software sectors are among the fastest adopters of AI coding tools, with widespread production use of AI-assisted coding and content management. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI code assistants (Copilot, Codeium, etc.) substantially raise developer productivity for drafting, refactoring, and testing updates while the developer retains control over architecture and validation, making this a strong augmentation case independent of full automation. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially speeds up developers' update workflows via code generation, automated testing suggestions, and content drafting, while humans retain control over deployment and quality assurance. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can automate routine updates (content replacement, minor styling changes, schema updates) and can generate code for standard components, but complex architectural changes, security-critical updates, and cross-system validation typically require human oversight, limiting time savings to roughly 40–60% depending on update complexity. |
| Task automatability | claude-sonnet-5 | 4/5 | Routine content and code updates (CMS edits, dependency bumps, minor UI tweaks) can largely be generated and applied by AI coding agents with human review, meeting the time-saving threshold for a large share of update work. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers prevent automation of updates; liability typically flows to the organization rather than a licensed individual, and organizational practices (code review, testing requirements) create friction but not hard barriers to AI-assisted or fully automated deployment pipelines. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or regulatory requirement mandates a human perform website updates; organizational risk tolerance is the main friction, not legal barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI coding assistance and automated deployment lower the marginal cost per routine update significantly, but oversight, integration, and human review for quality assurance keep total cost-per-task close to junior developer labor for non-trivial changes. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI-assisted update workflows drastically cut developer hours for routine changes, making inference plus lightweight review far cheaper than dedicated developer time for many update tasks. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (GitHub Copilot, various CI/CD automation systems, headless CMS platforms) reliably perform many classes of website updates in production; however, they still require human review for non-trivial logic, security implications, and validation against business requirements. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | AI coding assistants and agentic tools (e.g., Copilot, Cursor, CMS AI plugins) are deployed in production for code changes and content edits, but complex site updates involving design judgment or cross-system testing still require significant human oversight. |
Research, document, rate, or select alternatives for Web architecture or technologies.
61CI 56–66 · exposure 50 · augmentation 100 · importance 2.8/5 · click for rater detail
Research, document, rate, or select alternatives for Web architecture or technologies.
61| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Tech companies and digital-first organizations are actively deploying AI for documentation, code research, and decision support in engineering workflows. Adoption is measurable in production use of GitHub Copilot, ChatGPT in design reviews, and internal AI tools for architecture guidance, with clear upward trajectory in information sector. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector for AI tools, with widespread use of AI coding assistants and research tools already embedded in developer workflows. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI dramatically augments developer productivity for exploring architectural alternatives, generating comparison documents, and surfacing relevant technologies and patterns. Developers using AI-assisted research tools complete comprehensive evaluations substantially faster while maintaining quality oversight, making this a high-augmentation task. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI significantly accelerates the research and documentation phase of technology selection, surfacing comparisons, tradeoffs, and best practices quickly while the developer retains decision authority. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can assist significantly with research and documentation of web architectures and technologies, generating comparison matrices and technical summaries. However, the selection and rating of alternatives typically requires domain judgment about project constraints, team capabilities, and long-term maintenance—requiring human oversight to validate recommendations against organizational context. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can research and compare web technologies, summarize tradeoffs, and produce comparison tables, but final selection requires contextual judgment about organizational constraints, legacy systems, and team skill sets that AI cannot fully assess independently.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | No legal or licensing requirement mandates human sign-off, but organizational practice typically involves technical leads reviewing architecture recommendations. Risk aversion around technology selection decisions, combined with need for team alignment, creates moderate friction against full automation. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing or regulatory requirement governs choosing web technologies; this is a purely technical decision with no legal sign-off requirement. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference and integration costs for research, documentation, and initial filtering of alternatives are substantially lower than hiring senior engineers for comparative analysis. The per-task cost is roughly one-tenth of a developer's equivalent manual research time, though human oversight is still required. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI-assisted research and documentation of technology alternatives is very cheap compared to a developer spending hours reading docs and benchmarks, though a human is still needed to verify and finalize the analysis. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | LLM-based tools and AI code assistants can retrieve and summarize technology information reliably, but deployment for actual architecture decisions remains limited to assisted workflows. Mature products exist for documentation generation and tech research, but autonomous selection recommendations would require human verification before production adoption. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed tools like ChatGPT, Perplexity, and coding assistants (Copilot, Cursor) reliably help developers research and compare architectures, but they don't autonomously perform full evaluations without human framing and verification of claims. |
Create Web models or prototypes that include physical, interface, logical, or data models.
60CI 57–62 · exposure 50 · augmentation 88 · importance 3.2/5 · click for rater detail
Create Web models or prototypes that include physical, interface, logical, or data models.
60| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Web development and software engineering are digitally mature, high-velocity sectors with rapid AI tool adoption. Copilot, GitHub Copilot for Business, and ChatGPT-based workflows are mainstream in many organizations, and developers routinely use AI for code scaffolding and prototyping. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software/web development is a fast-adopting sector with widespread use of AI coding and design assistants in production workflows already. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly accelerates prototype generation, scaffolding, and boilerplate reduction while developers remain in control of architecture, validation, and refinement. This is one of the strongest AI-augmentation use cases in software development today. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI significantly speeds up prototyping and modeling by generating initial layouts, schemas, and code scaffolds, letting developers focus on refinement and integration. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can automate portions of prototype creation—generating boilerplate UI code, data schema templates, and basic interface layouts—but typically requires significant developer oversight, domain-specific customization, and integration with existing systems. The creative and architectural decisions that define a sound model still require human expertise. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can generate wireframes, data schemas, and interface prototypes from descriptions, but integrating these into a cohesive, context-aware model for a specific business still requires substantial human iteration and judgment. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing requirement or legal mandate for human sign-off exists; prototypes and models are often internal artifacts. Customer or organizational preference for human developers, and latent liability for flawed models, create some friction but not hard barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There are no licensing, legal, or regulatory requirements mandating human creation of web prototypes; adoption is purely a matter of organizational choice. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI inference is cheap per token, but the cost of integration, testing, architectural review, and human oversight—essential for model correctness—offsets raw inference cost. Overall cost is roughly comparable to a junior developer's time for quality output. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI tools cut prototyping time substantially and are cheap per-use, but human developer oversight, correction, and integration work still add meaningful cost, keeping the ratio moderate rather than order-of-magnitude cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Products like GitHub Copilot, ChatGPT with code generation, and design-assist tools exist and are deployed, but they produce code requiring substantial review and iteration. Error rates remain material for complex logical and data models, and reliability drops for non-standard or novel architectures. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Tools like v0, Figma AI, and code-generation assistants produce usable prototypes and scaffolding today, but they often require significant human refinement and are not fully autonomous for complex multi-model design. |
Recommend and implement performance improvements.
57CI 57–57 · exposure 50 · augmentation 88 · importance 3.4/5 · click for rater detail
Recommend and implement performance improvements.
57| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Web development is a digitally native, fast-moving sector with high AI tool adoption (Copilot, ChatGPT for coding, profilers); performance optimization is routine in agile workflows. Pilots and early production use are already common in tech-forward organizations. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector with widespread use of AI coding assistants and automated performance tooling integrated into CI/CD pipelines already in production. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI-powered profilers, linters, and code suggestion tools meaningfully accelerate a developer's ability to identify, test, and iterate on performance improvements, keeping them in decision-making and validation roles while radically speeding up the search-and-test loop. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools significantly speed up identifying bottlenecks, suggesting code-level fixes, and even generating optimized code snippets, meaningfully boosting developer productivity while humans validate and deploy changes. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can assist with identifying common performance bottlenecks (via code analysis tools) and suggest optimizations, but implementation requires human judgment on architecture trade-offs, context-specific constraints, and testing validation. Meaningful automation covers perhaps 40–50% of the workflow. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can identify common performance issues (e.g., bundle size, unoptimized images, render-blocking scripts) and suggest fixes, but implementing complex performance improvements across a live system requires profiling, testing, and judgment beyond current fully autonomous capability. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No licensing or regulatory requirement mandates human sign-off on performance improvements, and few contractual/liability barriers exist. Organizational inertia and developer preference for human-validated solutions provide modest friction, but nothing legally blocks substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but production system changes carry real risk of regressions, requiring code review and testing processes that create moderate organizational friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI tooling (subscription SaaS, inference) costs are now comparable to junior developer time for routine optimizations, but senior expertise in performance engineering commands higher rates; all-in costs roughly trade even on average tasks. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted profiling and code suggestions reduce developer time somewhat, but the need for human review, testing, and deployment oversight keeps costs roughly comparable to a developer doing this with AI assistance rather than order-of-magnitude cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Products like GitHub Copilot, static analysis tools, and AI-powered profilers exist and can detect and suggest improvements in production, but they produce false positives, miss domain-specific issues, and require significant developer review and override—not fully autonomous end-to-end performance tuning. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Tools like Lighthouse, AI-assisted linters, and coding assistants (Copilot, Cursor) can flag and even patch performance issues, but reliable end-to-end performance optimization in production systems still needs human validation and testing. |
Develop databases that support Web applications and Web sites.
52CI 38–66 · exposure 38 · augmentation 88 · importance 4.0/5 · click for rater detail
Develop databases that support Web applications and Web sites.
52| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Tech companies and development teams are adopting AI for coding tasks at moderate velocity, including database-related work, but mostly as assistive tools. AI is used to speed up junior developers' schema drafting rather than replace senior database architects in production decision-making. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is among the fastest AI-adopting fields, with AI coding tools now embedded in mainstream developer workflows including database-related tasks like schema generation and query writing. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists web developers by generating schema templates, suggesting indexing strategies, auto-completing SQL, and catching common design errors. Developers remain in the loop for final validation and optimization, but AI transforms productivity on routine and repetitive aspects of database development. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools substantially speed up writing schemas, queries, migrations, and boilerplate ORM code, letting developers focus on architecture and optimization while AI handles repetitive generation work. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Database development requires significant architectural decisions, schema design, and integration logic tailored to specific application needs. While AI can assist with boilerplate SQL/schema generation and suggest patterns, end-to-end database development demands domain expertise, performance optimization, and security considerations that current systems cannot reliably handle autonomously at 50% time savings with equal quality. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can generate schema designs, SQL/NoSQL queries, and migrations from specifications, but real database development involves architecture decisions, performance tuning, and iterative debugging that still need human oversight, limiting full end-to-end automation today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Database decisions carry high downstream liability and performance consequences; organizations typically retain human architects and DBAs to validate and sign off on database designs. Professional responsibility norms and the need for expert oversight create meaningful friction to full automation. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, legal, or regulatory barriers restrict who can design or build databases; this is a purely technical task with no human-contact or authorization requirement. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI-generated database systems that require significant expert review, redesign, and debugging approaches or exceeds the cost of direct human development. Integration and oversight overhead is substantial for a task where errors compound across the entire application. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | AI-assisted schema and query generation is very cheap compared to a developer's hourly cost for routine database setup tasks, though oversight and correction for complex systems still add human cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI tools can generate database schemas and basic SQL, but deployed products lack reliable performance for complex database design, normalization decisions, and production-scale optimization. Most AI assistance is at the code-snippet level rather than full database architecture development that meets real application requirements. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Coding assistants (Copilot, Cursor, Claude) and AI-powered ORMs reliably generate schema scaffolding and boilerplate queries in production use, but complex database architecture, indexing strategy, and scaling decisions still require significant human expertise and review. |
Evaluate or recommend server hardware or software.
47CI 36–59 · exposure 38 · augmentation 75 · importance 3.1/5 · click for rater detail
Evaluate or recommend server hardware or software.
47| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Tech and cloud-native companies are actively using AI-assisted code generation and infrastructure-as-code tools, but hardware/software evaluation recommendations are less frequently fully automated; adoption is growing in DevOps contexts but remains mixed across the development sector. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Software/IT sectors adopt AI coding and research tools quickly, but infrastructure decision-making remains a slower-adopted, judgment-heavy niche within that fast-moving field. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at synthesizing vendor specifications, generating comparison tables, and surfacing performance/cost trade-offs, significantly accelerating a developer's evaluation workflow while the human retains judgment on fit, risk, and organizational constraints. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI is very useful for quickly summarizing hardware/software options, benchmarks, and tradeoffs, substantially speeding up the research phase even though final judgment stays human. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can perform technical research, compare specifications, and generate recommendations based on published benchmarks and requirements analysis, but evaluation of server hardware/software typically requires domain-specific context (workload patterns, budget constraints, organizational constraints) and risk assessment that often requires human judgment for final sign-off. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires synthesizing organizational constraints, budget, scaling needs, and vendor tradeoffs; AI can supply research and comparisons but cannot reliably own the recommendation end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Recommendation authority often rests with human architects or senior developers who must take responsibility for performance/cost/risk outcomes; organizations typically require human judgment sign-off before procurement, creating adoption friction even if AI can generate the technical analysis. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No formal licensing is required, but decisions carry real cost/liability implications (downtime, security, budget), so organizations typically want an accountable human sign-off. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference for analyzing specifications and generating comparison matrices is very cheap (pennies per task), while a web developer's time for this evaluation costs $50–$150/hr; AI-generated recommendations require modest human review overhead, making the cost ratio heavily favor automation. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI can cheaply generate research and draft comparisons, but a human architect must still validate and integrate context-specific factors, so overall cost savings are moderate rather than dramatic. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed AI systems (coding assistants, cloud platform recommendation engines) can produce technically sound hardware/software suggestions, but they struggle with nuanced contextual factors like legacy system compatibility, organizational politics, and long-term maintenance costs that affect real-world recommendations. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI assistants can produce comparison tables and technical summaries but no deployed product independently evaluates and recommends infrastructure decisions for organizations without heavy human vetting. |
Select programming languages, design tools, or applications.
37CI 25–49 · exposure 25 · augmentation 75 · importance 4.2/5 · click for rater detail
Select programming languages, design tools, or applications.
37| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While tech companies experiment with AI-assisted tool suggestions in IDEs and chat systems, actual adoption of AI-driven autonomous technology selection in production is minimal. Most organizations still rely on architect-led decisions and RFP processes, with only pilots underway in early-adopter firms. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector where AI coding assistants (Copilot, ChatGPT, etc.) are already widely used for architecture and stack recommendations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at augmenting this task by rapidly generating pros/cons, comparing frameworks, providing code examples, and surfacing community patterns to help developers make faster, more informed decisions. The human architect or lead retains decision authority while AI significantly accelerates research and analysis. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI is highly useful for comparing languages/frameworks, summarizing tradeoffs, and speeding up research, significantly augmenting developer decision-making even though final selection remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can suggest programming languages or tools based on project requirements, but selecting the optimal stack requires understanding business constraints, team expertise, legacy systems, and architectural tradeoffs that demand human judgment. The task cannot meet the 50%-time-saving bar as a full end-to-end automation. |
| Task automatability | claude-sonnet-5 | 2/5 | This is a judgment-heavy architectural decision involving tradeoffs across team skill, scalability, maintainability, and business context; AI can suggest options but cannot reliably make the final call end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Technology selection is typically a governance decision that involves architectural review boards, compliance teams, and organizational oversight in enterprises. Team leads or senior engineers must sign off, creating organizational and accountability barriers that prevent autonomous AI substitution. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, regulatory, or liability barriers prevent using AI to inform or assist this decision. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | An AI tool that provides suggestions still requires a human senior engineer to validate and make the final decision, so the total cost remains dominated by the expert human. The AI cost savings are marginal compared to the loaded wage of a qualified developer making this decision. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | Consulting an AI for recommendations is cheap and fast, but the human still must evaluate, research, and validate the decision, so overall cost savings are moderate rather than dramatic. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI systems can generate recommendations through prompt-based interfaces or chatbots, no deployed product reliably makes production-grade technology selections across complex organizational contexts. Products exist in a narrow advisory capacity but lack the contextual reasoning for autonomous, high-stakes technology decisions. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI coding assistants can recommend languages/frameworks when prompted, but no deployed product autonomously selects and commits to a tech stack for a project without heavy human vetting. |
Analyze user needs to determine technical requirements.
35CI 32–38 · exposure 25 · augmentation 75 · importance 3.8/5 · click for rater detail
Analyze user needs to determine technical requirements.
35| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | AI tools are increasingly used in software development environments (GitHub Copilot, ChatGPT in workflows), but requirements analysis remains predominantly human-driven; adoption is rising but still pilot-heavy rather than production-default. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Software/professional services sectors adopt AI tools quickly for drafting and analysis support, but full requirements-gathering automation remains uncommon in production workflows. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at augmenting this task: developers use LLMs to brainstorm requirements, generate candidate specifications, and summarize stakeholder feedback, materially accelerating the analysis phase while developers retain final judgment and client alignment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can effectively assist by summarizing stakeholder input, generating draft requirement docs, and suggesting technical specifications, meaningfully speeding up the analyst's work. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in parsing structured data and suggesting technical requirements, analyzing user needs involves nuanced stakeholder communication, context understanding, and judgment that typically requires human expertise to reach 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Requirements gathering depends heavily on stakeholder interviews, ambiguous business context, and judgment calls that current AI cannot reliably substitute end-to-end.assist in drafting but not fully replace this task. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Client sign-off on requirements is typically a business and contractual expectation rather than a legal requirement, and many organizations prefer human ownership of requirements; however, no hard licensing barrier prevents AI assistance. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational trust, client relationship needs, and accountability for correctly capturing business needs create moderate friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI systems (LLM APIs, coding assistants) cost little per inference but require substantial human review and iteration to produce usable requirements, making the true all-in cost closer to or exceeding the cost of direct human analysis. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Because human oversight and clarification loops with clients/stakeholders remain necessary, cost savings are limited despite some drafting efficiency gains. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No mainstream product reliably performs end-to-end needs analysis autonomously; AI tools can draft requirement documents or suggest features from transcripts, but they frequently miss domain-specific context and require significant human revision. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools (chatbots, requirement-analysis assistants) exist but are narrow, often producing generic requirement drafts that need heavy human validation and clarification with stakeholders. |
Develop or implement procedures for ongoing Web site revision.
35CI 32–38 · exposure 25 · augmentation 75 · importance 3.3/5 · click for rater detail
Develop or implement procedures for ongoing Web site revision.
35| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | The software/tech sector actively pilots and deploys AI coding assistants, but autonomous procedure development remains rare in production; most adoption is still in the assistance phase rather than full automation. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Web development and tech sectors show fast AI tool adoption generally, but procedural/process-design work specifically lags behind code-generation use cases. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI coding assistants significantly boost developer productivity by auto-generating boilerplate, suggesting improvements, and drafting documentation, while developers retain control over procedure architecture and business logic integration. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can effectively help draft revision procedures, changelogs, style guides, and workflow documentation, meaningfully speeding up the human-led process design. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with code generation and static content updates, developing comprehensive revision procedures requires understanding business logic, user requirements, testing protocols, and deployment strategies that typically demand human judgment and context. Only narrow sub-tasks (e.g., routine code formatting) could be fully automated without significant domain input. |
| Task automatability | claude-sonnet-5 | 2/5 | Developing governance procedures for ongoing site revision requires organizational judgment, stakeholder alignment, and process design that AI cannot fully execute end-to-end today, though it can assist with drafting components. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | Web development changes are often subject to organizational governance, version control policies, and stakeholder approval; there is friction around autonomous procedure design, though no strict legal licensing barrier to automation itself. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizational processes typically need buy-in from stakeholders and IT governance, creating moderate friction against pure AI-driven implementation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI coding assistants reduce some drafting costs but do not approach the cost of a full developer salary when all overhead, integration, and review oversight are factored in. The human still directs the revision strategy and validates outputs. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Because the task involves cross-team coordination and decision-making, AI reduces some drafting time but still requires substantial human oversight, keeping costs closer to comparable rather than dramatically cheaper. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | Some AI tools can generate code snippets and suggest refactoring, but no deployed product reliably develops or implements full revision procedures end-to-end—doing so requires integrating with CI/CD systems, stakeholder alignment, and change management decisions that remain primarily human-driven in production. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | No deployed product autonomously creates and implements website revision workflows; AI coding assistants help with content/version updates but not the broader procedural/process design work. |
Communicate with network personnel or Web site hosting agencies to address hardware or software issues affecting Web sites.
34CI 30–38 · exposure 25 · augmentation 75 · importance 3.6/5 · click for rater detail
Communicate with network personnel or Web site hosting agencies to address hardware or software issues affecting Web sites.
34| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Web hosting and infrastructure support remain relationship-heavy and context-specific; adoption of full AI automation here is slow because of trust, liability, and the need for human judgment in escalation decisions. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Web development and IT operations are digitized and fast-adopting AI tools (e.g., for log analysis, incident triage), but full communication handling with external agencies remains a human-led process in most orgs today. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI excels at drafting clear incident summaries, suggesting relevant technical documentation, and preparing diagnostic information—substantial assistance that helps developers communicate more efficiently with network teams while the developer retains full decision authority. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly helps by drafting technical support messages, summarizing error logs, suggesting fixes, and accelerating diagnosis, letting the developer focus on judgment calls and vendor relationships. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | AI can draft communications and suggest technical solutions, but the task requires real-time troubleshooting dialogue, coordination across multiple external parties, and judgment about which issues to escalate—human involvement remains essential for most interactions. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires real-time troubleshooting dialogue, diagnosing ambiguous infrastructure issues, and coordinating with third parties, which involves judgment and back-and-forth that AI can support but not fully replace end-to-end today.communication also often requires contextual authority and accountability. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While communication itself has no legal barrier, the task sits within SLAs and service agreements that typically require a human representative's accountability; customer contracts often mandate human technical contact for issue resolution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No strict licensing barrier, but organizational friction exists since resolving hosting/network issues often requires accountability, escalation paths, and trust relationships that favor human communication. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted communication tools (templates, draft generation) reduce overhead, but the specialized expertise and relationship-building required mean a human Web developer's cost remains lower than AI-only alternatives for critical infrastructure coordination. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can assist in drafting diagnostic messages or interpreting logs cheaply, the need for human verification, escalation, and relationship management with vendors keeps the effective cost comparable to human labor. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While LLMs can generate template emails and chatbots handle routine inquiries, no mature product reliably manages complex, context-dependent communication with external technical teams involving live hardware/software diagnostics without human oversight. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI chatbots and copilots can help diagnose common server/DNS/hosting errors, but production systems don't reliably handle the full communication and negotiation loop with hosting agencies or network teams without human oversight. |
Incorporate technical considerations into Web site design plans, such as budgets, equipment, performance requirements, and legal issues including accessibility and privacy.
32CI 28–36 · exposure 25 · augmentation 75 · click for rater detail
Incorporate technical considerations into Web site design plans, such as budgets, equipment, performance requirements, and legal issues including accessibility and privacy.
32| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Web development sectors show moderate AI adoption in code generation and drafting tools, but strategic planning and requirements synthesis remain largely human-driven. Pilots of AI-assisted compliance checking exist, but production-scale adoption of AI-owned planning is not yet widespread. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Web development is a digitized, fast-moving field with growing AI tool adoption for coding and design assistance, though the planning/compliance-synthesis aspect lags behind pure code generation adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can meaningfully assist by auto-generating accessibility audit checklists, flagging common privacy pitfalls, summarizing budget constraints, and drafting technical requirement documentation. These tools speed up the human developer's planning process substantially while keeping the developer in control of final decisions. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by generating accessibility guidelines, summarizing legal requirements, estimating performance benchmarks, and drafting technical specifications, substantially speeding up the planning process while a human finalizes decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in drafting accessibility and privacy documentation, and can identify some technical constraints, the task fundamentally requires human judgment to weigh competing priorities (budget vs. performance vs. legal risk) and make design trade-offs that depend on business context and client goals. Current systems cannot reliably translate complex, context-dependent requirements into coherent design plans. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires synthesizing budget constraints, equipment/infrastructure choices, performance targets, and legal compliance into a coherent plan, which involves organizational judgment and stakeholder negotiation AI cannot fully replicate end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Accessibility and privacy compliance involve legal and regulatory obligations (WCAG, GDPR, CCPA, etc.) that typically require human professional accountability. Liability asymmetry is high: an AI-generated plan that misses a legal exposure creates organizational risk, so human sign-off and expertise remain quasi-mandatory. |
| Adoption barriers | claude-sonnet-5 | 3/5 | Legal issues like accessibility (ADA/WCAG) and privacy (GDPR/CCPA) carry liability exposure, creating moderate barriers since organizations want accountable humans reviewing compliance-related design decisions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 3/5 | AI-assisted compliance and requirements gathering could reduce labor cost moderately, but human developers must still validate, reconcile, and own the final design plan. The cost-per-output is roughly comparable to human-only approaches given the oversight required. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can cheaply generate draft checklists or compliance summaries, the actual planning requires human review of legal risk and business constraints, keeping all-in costs closer to human-comparable. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs this end-to-end task. AI can generate compliance checklists and suggest technical considerations, but real-world design planning involves negotiation, prioritization, and integration of multifaceted business and legal constraints that current systems handle only superficially or with high error rates. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools can draft accessibility checklists or suggest performance benchmarks, but no deployed product reliably integrates budget, legal, and technical planning into a finalized site design plan without heavy human oversight. |
Design and implement Web site security measures, such as firewalls and message encryption.
29CI 25–32 · exposure 25 · augmentation 63 · click for rater detail
Design and implement Web site security measures, such as firewalls and message encryption.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Security roles remain among the most human-centric in tech; organizations are slow to automate security design and architecture decisions because errors have high consequences. Adoption of AI for this task is limited to narrow, low-risk code-generation assist rather than full system design. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Software/tech sectors are fast adopters of AI coding tools, but security-specific automation lags due to risk sensitivity, placing this in middling production adoption. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can usefully assist by generating boilerplate encryption code, suggesting firewall rules based on patterns, and identifying potential vulnerabilities, improving a security engineer's productivity on routine implementation tasks. However, augmentation is limited because the core work—threat modeling and architectural decisions—remains fundamentally human. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI coding assistants substantially speed up writing encryption code, configuring firewalls, and identifying common vulnerabilities, meaningfully boosting developer productivity while humans retain final design responsibility. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Security design requires expert judgment about threat models, architecture trade-offs, and organizational context that current AI cannot reliably perform end-to-end. While AI can assist with boilerplate encryption implementation and firewall rule generation, the critical decisions about which measures to deploy and how to integrate them demand human expertise and cannot achieve 50% time savings at equal quality autonomously. |
| Task automatability | claude-sonnet-5 | 2/5 | Designing and implementing security architecture requires contextual judgment about threat models, infrastructure, and compliance that current AI cannot fully replace, though AI can help draft configs and identify vulnerabilities.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Security implementations carry substantial liability and regulatory requirements; incorrect firewalls or encryption can expose an organization to legal and compliance breach. In regulated industries, a qualified human must typically sign off on or legally implement critical security measures, creating strong organizational and legal barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement specifically for this task, but liability for security breaches and compliance obligations (PCI-DSS, GDPR) create strong organizational caution against fully automating it. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI assistance for security tasks still requires senior engineers to review, validate, and take responsibility for implementations, meaning the human labor cost remains dominant. Full autonomy is not achievable, so cost savings are marginal compared to the loaded wage of a security-competent developer. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Security design still requires expert human oversight to avoid costly breaches, so AI assistance reduces some labor but doesn't yet substantially undercut the cost of a skilled security-conscious developer. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed AI product reliably designs and implements complete security architectures independently. Tools exist for code generation and vulnerability scanning, but they require substantial human oversight, threat assessment, and architectural decision-making to ensure security adequacy. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Products like Copilot or security scanners assist with parts of this (suggesting firewall rules, flagging vulnerabilities) but no deployed product independently designs and implements full security architectures reliably. |
Maintain understanding of current Web technologies or programming practices through continuing education, reading, or participation in professional conferences, workshops, or groups.
28CI 16–40 · exposure 17 · augmentation 75 · importance 3.8/5 · click for rater detail
Maintain understanding of current Web technologies or programming practices through continuing education, reading, or participation in professional conferences, workshops, or groups.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for this meta-task (learning about learning) is nascent; while generative tools assist with content summarization, actual displacement of developer time spent in continuing education is minimal because the task involves intentional professional growth that organizations expect humans to own. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Software development is a fast-adopting sector, and developers widely use AI tools (chatbots, coding assistants, curated feeds) to keep up with new technologies. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments this task by summarizing conference talks, curating relevant articles, generating study guides, and filtering news feeds, meaningfully raising a developer's productivity in staying current without the developer ceding agency over what to learn. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly boosts a developer's ability to learn new tools quickly by summarizing documentation, explaining changes, and answering technical questions, greatly aiding continuous learning. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task fundamentally requires human judgment about which emerging technologies are relevant, critical evaluation of content, and intentional professional development choices. AI cannot autonomously decide what a developer needs to learn or replace the reflective learning process itself. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can summarize articles, curate learning resources, and explain new frameworks, but the underlying task requires personal ongoing learning, professional networking, and judgment that isn't a discrete deliverable AI can fully replace. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Professional development and staying current are primarily human responsibilities embedded in organizational culture, career advancement expectations, and individual autonomy; mandates and licensing do not strictly prevent automation, but the inherent need for human agency in choosing what to learn and engaging with communities creates strong structural barriers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal barrier prevents AI-assisted learning, but the task is intrinsically tied to individual professional development and networking, limiting full substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted curation tools exist but are inexpensive only for content summarization; they do not replace the time humans must spend actually learning, attending conferences, or networking—tasks that require human time regardless of automation. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply supply summaries and explanations, but since the task is inherently personal and ongoing, there's no direct human-labor cost being replaced at scale to compare against. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While AI can summarize tech news, curate articles, or generate reading lists, no deployed product reliably replaces the human's active engagement with continuing education or professional conferences, which inherently require human presence and decision-making. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI chat tools and news aggregators exist to help developers stay current, but no deployed product autonomously performs continuing education or conference participation on a developer's behalf. |
Collaborate with management or users to develop e-commerce strategies and to integrate these strategies with Web sites.
26CI 14–38 · exposure 13 · augmentation 75 · importance 3.5/5 · click for rater detail
Collaborate with management or users to develop e-commerce strategies and to integrate these strategies with Web sites.
26| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Web development teams are moderately adopting AI for documentation and code generation, but strategic collaboration remains human-centric. Pilots of AI-assisted strategy exist in information-sector firms, but production adoption of autonomous strategy development is limited. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Web development and e-commerce sectors have moderate-to-fast AI tool adoption for coding and content generation, though strategic consulting work adopts more slowly. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists this task by generating strategy outlines, competitive analysis, integration recommendations, and documentation drafts that humans then refine. These tools measurably raise the productivity of strategy discussions and reduce iteration time while keeping human judgment central. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by analyzing market trends, generating strategy drafts, and prototyping site integrations, significantly speeding up the human-led strategic process. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with strategy drafting and documentation, this task requires deep understanding of business goals, user needs, and organizational context that humans must ultimately decide. The collaborative and decision-making components are not automatable end-to-end; AI cannot replace the human-led strategy development process. |
| Task automatability | claude-sonnet-5 | 1/5 | This task centers on interactive strategic collaboration with stakeholders requiring judgment, negotiation, and business context that current AI cannot autonomously replicate end-to-end.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | E-commerce strategy must align with business objectives, legal compliance, and user protection—areas where accountability and liability rest with human decision-makers. Management sign-off and user validation are quasi-regulatory requirements; organizations cannot fully delegate strategy approval to AI. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing barrier, but organizational trust, stakeholder relationship management, and accountability for strategic decisions create moderate friction against full AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | While AI drafting and analysis tools reduce some overhead, the core task requires significant human expertise and oversight. The all-in cost (including management review, validation, and integration work) remains high relative to AI-only approaches; human strategists remain the primary cost driver. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI can cheaply generate analysis or draft content, the human-facing strategic negotiation portion still requires costly skilled labor, keeping overall cost comparable to human-only delivery. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No mature product reliably conducts end-to-end e-commerce strategy collaboration and integration at production scale. AI tools can generate strategy suggestions or help document decisions, but deployed systems do not independently conduct stakeholder management or validate strategy fit with organizational constraints. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently develops e-commerce strategy through management collaboration; AI tools only assist in drafting or analysis subtasks. |
Confer with management or development teams to prioritize needs, resolve conflicts, develop content criteria, or choose solutions.
22CI 11–32 · exposure 13 · augmentation 75 · importance 3.6/5 · click for rater detail
Confer with management or development teams to prioritize needs, resolve conflicts, develop content criteria, or choose solutions.
22| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Adoption of AI for core team conferencing and conflict resolution is minimal; most organizations still view these as distinctly human leadership functions. While some teams use AI for meeting transcription and analysis, few are replacing human-led prioritization and conflict resolution. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Tech/software teams are fast adopters of AI tools generally, but for facilitation and conflict resolution specifically, adoption remains at the assistive-tool stage rather than replacement. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can meaningfully assist by analyzing requirements, generating decision matrices, summarizing meeting context, and surfacing potential conflicts before human discussion—all of which can accelerate and improve human-led conferencing. However, humans retain responsibility for final decisions and relationship management. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by summarizing requirements, drafting content criteria, analyzing trade-offs, and preparing briefing materials that inform the human-led conversation. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with brainstorming solutions and drafting decision frameworks, the core task requires human judgment on prioritization, conflict resolution, and stakeholder negotiation. Current AI cannot reliably mediate team conflicts or make final priority decisions that account for unmeasured organizational context and politics. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires synthesizing organizational politics, stakeholder relationships, and real-time negotiation among humans with competing interests, which current AI cannot autonomously conduct end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Team leadership and management responsibilities typically require human authority and accountability; organizations have strong cultural norms against ceding final prioritization decisions to AI systems. Liability for bad prioritization decisions and the need for human ownership create substantial organizational barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No formal licensing requirement, but organizational trust, accountability for decisions, and interpersonal dynamics create real friction against full AI substitution. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | The cost of AI-generated meeting notes and decision support is relatively low, but the task itself is typically performed by senior developers or architects earning substantial wages. Full replacement would require AI to handle all interactive and judgment-based elements, making net cost savings uncertain. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Human facilitators, managers, and developers still must be present and paid for their time; AI tools reduce prep/documentation costs only marginally. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs end-to-end conferencing and conflict resolution. AI tools can generate meeting summaries and suggest decision frameworks, but cannot substitute for the human interactive dialogue, consensus-building, and real-time conflict navigation this task requires. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product runs cross-team conflict resolution or prioritization meetings independently; AI is at best a note-taker or summarizer in these settings. |
Related occupations — Computer & Mathematical
How to read this
A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.
What would change this score
New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.