Computer Systems Engineers/Architects

15-1299.08
Median wage $116,580/yr435,370 employed (US)Rank #114 of 923 scored · top 12% by substitution

Design and develop solutions to complex applications problems, system administration issues, or network concerns. Perform systems management and integration functions.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution45
Exposure41
Augmentation80

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

28 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

7%

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%41

panel mean rating 2.6/5 → substitution pressure 41/100

Technical feasibility todayw 20%42

panel mean rating 2.7/5 → substitution pressure 42/100

Cost vs. human wagew 15%44

panel mean rating 2.8/5 → substitution pressure 44/100

Adoption barriersw 20%inverted — strong barriers lower the score53

panel mean rating 2.9/5 (barrier strength) → substitution pressure 53/100

Sector adoption velocityw 10%56

panel mean rating 3.2/5 → substitution pressure 56/100

Task breakdown (28 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.

Monitor system operation to detect potential problems.

81

CI 66–95 · exposure 80 · augmentation 100 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Monitoring automation has achieved deep, rapid adoption across tech, finance, and cloud-native sectors; nearly all enterprises rely on automated monitoring as standard infrastructure practice.
Sector adoption velocityclaude-sonnet-54/5IT operations and cloud infrastructure sectors have rapidly adopted AI-driven monitoring and AIOps tools, reflecting fast adoption patterns typical of technology-heavy fields.
Augmentation potentialclaude-haiku-4-5-202510015/5AI monitoring systems routinely augment engineer productivity by surfacing prioritized alerts, correlating root causes, and recommending remediations, enabling engineers to act on intelligence rather than manually scanning dashboards.
Augmentation potentialclaude-sonnet-55/5AI substantially amplifies an engineer's ability to monitor complex distributed systems by surfacing anomalies, correlating logs, and predicting failures well beyond manual capacity.
Task automatabilityclaude-haiku-4-5-202510015/5System monitoring for anomaly detection is highly automatable; ML-based monitoring tools and agents can continuously analyze logs, metrics, and alerts 24/7, detecting pattern deviations far faster than humans and delivering >50% time savings with equal or superior quality.
Task automatabilityclaude-sonnet-53/5Automated monitoring, anomaly detection, and alerting tools already handle much of the detection work, but interpreting root causes and deciding on system-wide architectural implications still requires human judgment for complex systems.
Adoption barriersclaude-haiku-4-5-202510012/5While some organizations prefer human oversight layers, there are no licensing or legal barriers preventing automation; monitoring is typically internal and governed only by organizational risk tolerance and SLA requirements.
Adoption barriersclaude-sonnet-52/5No licensing requirement for monitoring, but critical infrastructure oversight often retains human sign-off for major system changes triggered by detected problems, creating mild organizational friction.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI-based monitoring costs per monitored system (subscription + overhead) are typically orders of magnitude cheaper than employing full-time human monitoring staff or on-call engineers dedicated to the same task.
Cost vs. human wageclaude-sonnet-54/5Automated monitoring tools are inexpensive relative to constant human observation and scale across many systems simultaneously, though initial integration and tuning costs exist.
Technical feasibility todayclaude-haiku-4-5-202510015/5Deployed products like Datadog, New Relic, Splunk, and cloud-native monitoring agents reliably perform continuous system monitoring in production across millions of systems globally, with mature alerting and anomaly detection.
Technical feasibility todayclaude-sonnet-54/5Mature APM/observability platforms (Datadog, New Relic, Splunk, AI-driven anomaly detection) are widely deployed in production and reliably flag issues, though they still generate false positives and need human triage for architecture-level problems.

Document design specifications, installation instructions, and other system-related information.

74

CI 70–79 · exposure 70 · augmentation 100 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Tech and software engineering sectors are among the earliest and fastest adopters of AI for documentation, code generation, and related tasks. Evidence shows widespread use in development teams, DevOps, and SRE workflows, with integration into CI/CD pipelines and knowledge management systems accelerating.
Sector adoption velocityclaude-sonnet-54/5Software/IT sectors are fast adopters of AI coding and documentation tools, with widespread use of AI assistants in engineering workflows already.
Augmentation potentialclaude-haiku-4-5-202510015/5AI significantly amplifies engineer productivity by automating first-draft generation, keeping humans focused on verification, architecture decisions, and nuanced explanations. This assistive pattern is already mature and widely adopted in development environments.
Augmentation potentialclaude-sonnet-55/5AI substantially speeds up drafting, formatting, and summarizing technical documentation while engineers retain responsibility for accuracy and final content.
Task automatabilityclaude-haiku-4-5-202510014/5AI can generate substantial portions of technical documentation from code, architecture diagrams, and specifications with high quality and significant time savings. While some domain-specific detail and review cycles remain, AI systems today can handle end-to-end drafting of standard sections like installation guides, API documentation, and system overviews, meeting the ≥50% time-saving threshold.
Task automatabilityclaude-sonnet-54/5Drafting documentation from existing design artifacts, code, or engineer notes is a well-suited generative AI task; LLMs can produce structured specs and installation guides quickly, though final review is needed for accuracy.
Adoption barriersclaude-haiku-4-5-202510012/5Documentation is not legally gatekept; no certification or signature requirement applies. The primary friction is organizational QA processes and the preference for human review of accuracy-critical sections, but these are soft procedural barriers rather than hard regulatory ones.
Adoption barriersclaude-sonnet-52/5No licensing requirement for writing internal documentation, though some regulated industries require engineer sign-off on specs, creating mild friction.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI documentation generation costs a fraction of a cent per output, while a systems engineer's time costs $50–100+ per hour. Even accounting for oversight and refinement, the cost differential is 100–1000x in AI's favor for routine documentation tasks.
Cost vs. human wageclaude-sonnet-54/5AI-assisted drafting is dramatically cheaper per page than a systems engineer manually writing documentation, even accounting for review time.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed products (GitHub Copilot, Claude, ChatGPT) reliably generate technical documentation in production settings, particularly for standard formats and common architectures. Minor gaps in domain specificity and occasional hallucinations exist, but the technology is proven at scale in professional environments.
Technical feasibility todayclaude-sonnet-53/5Products like Copilot, ChatGPT, and documentation-generation tools are used in production to draft technical docs, but they still require human verification for correctness and completeness of system-specific details.

Research, test, or verify proper functioning of software patches and fixes.

65

CI 55–75 · exposure 62 · augmentation 75 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5High-tech, digitized sectors (software, cloud, finance, tech services) have rapidly adopted automated patch testing and CI/CD verification; this represents mainstream production practice in large organizations.
Sector adoption velocityclaude-sonnet-53/5Software engineering and IT sectors show above-average AI tool adoption for code review and testing, though full automation of patch verification remains at the pilot stage in most organizations.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly augments engineers by automating test execution, generating test cases, and flagging anomalies, allowing human engineers to focus on interpreting results and strategic validation decisions rather than manual testing.
Augmentation potentialclaude-sonnet-54/5AI tools meaningfully speed up test generation, log analysis, and anomaly detection during patch verification, letting engineers focus on judgment calls while staying in the loop.
Task automatabilityclaude-haiku-4-5-202510014/5AI can automate large portions of patch testing through automated test generation, execution, and result analysis, achieving significant time savings. However, verification of complex system interactions and edge cases often still requires human oversight, preventing a full 5-rating.
Task automatabilityclaude-sonnet-53/5AI can generate test cases, run automated test suites, and analyze patch diffs for potential issues, but verifying proper functioning often requires domain context, integration testing, and judgment about edge cases that current systems only partially handle.:
Adoption barriersclaude-haiku-4-5-202510012/5Minimal regulatory or legal barriers exist for automating patch testing; organizations largely control their own testing protocols. Some organizational inertia around tool adoption and need for human sign-off on critical patches create modest friction.
Adoption barriersclaude-sonnet-52/5No licensing requirement blocks AI from assisting with this task, though organizational risk tolerance for unverified changes to production systems creates some friction and preference for human sign-off.
Cost vs. human wageclaude-haiku-4-5-202510014/5Automated testing and verification are orders of magnitude cheaper than manual human testing when deployed at scale, with marginal inference and integration costs compared to loaded engineer wages for exhaustive manual patch validation.
Cost vs. human wageclaude-sonnet-53/5AI-assisted testing can reduce time spent on repetitive verification tasks, but the need for human review of results and edge-case handling keeps overall cost savings moderate rather than order-of-magnitude.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed products (CI/CD platforms, automated testing frameworks, LLM-assisted test generators) reliably perform patch testing and verification in production environments at scale. Some gaps remain in complex integration scenarios and novel failure modes.
Technical feasibility todayclaude-sonnet-53/5Products like AI-assisted testing tools, CI/CD pipeline integrations, and code review assistants exist and are used in production, but full autonomous verification of patches without human oversight is not yet standard practice.

Configure servers to meet functional specifications.

64

CI 32–95 · exposure 62 · augmentation 75 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Information technology and cloud services sectors show deep, rapid adoption of automated infrastructure provisioning and configuration management; this is already standard practice at major tech firms and is spreading to enterprise IT across all sectors.
Sector adoption velocityclaude-sonnet-53/5IT/software sectors are moderately fast adopters of AI coding and DevOps tools, with growing use of AI-assisted infrastructure-as-code, though full autonomous server configuration remains uncommon in production.
Augmentation potentialclaude-haiku-4-5-202510014/5AI tools assist engineers by auto-generating configuration boilerplate, validating syntax, detecting misconfigurations, and suggesting optimizations, significantly raising productivity while engineers retain oversight of critical decisions and security policies.
Augmentation potentialclaude-sonnet-54/5AI coding assistants meaningfully speed up writing configuration scripts, generating templates, and troubleshooting syntax, significantly boosting engineer productivity while they retain final control.
Task automatabilityclaude-haiku-4-5-202510015/5Configuration of servers to meet functional specifications is highly codifiable and repetitive, involving deployment scripts, infrastructure-as-code tools (Terraform, Ansible), and automated provisioning. Current AI systems can generate, validate, and execute these configurations end-to-end with >50% time savings compared to manual setup.
Task automatabilityclaude-sonnet-52/5Server configuration requires understanding specific infrastructure context, dependencies, and functional requirements that vary widely; AI can assist with scripts/config templates but full end-to-end configuration with equal quality is not yet reliable without human oversight.
Adoption barriersclaude-haiku-4-5-202510012/5Minimal legal or licensing barriers exist; the primary friction is organizational (change management, testing requirements, security sign-off) and technical debt, but nothing prevents automation or substitution of the core task.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but organizational risk aversion around production infrastructure misconfiguration, security compliance, and change management processes create real friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510015/5Automated configuration infrastructure costs pennies per deployment after setup, while engineer labor for manual server configuration costs tens to hundreds of dollars per instance; AI-driven systems are orders of magnitude cheaper at scale.
Cost vs. human wageclaude-sonnet-52/5AI can reduce time on boilerplate config scripting, but the integration, validation, and troubleshooting still require skilled engineer time, keeping overall cost comparable to human-led work in most cases.
Technical feasibility todayclaude-haiku-4-5-202510015/5Mature products exist and are deployed at scale: infrastructure-as-code platforms, cloud provider automation services (AWS CloudFormation, Azure Resource Manager), and AI-assisted code generation tools reliably perform server configuration in production environments across major organizations.
Technical feasibility todayclaude-sonnet-52/5Tools like AI-assisted IaC generators (Terraform, Ansible playbook suggestions via Copilot) exist but are narrow-scope aids rather than autonomous, reliable configuration systems deployed at scale in production.

Develop system engineering, software engineering, system integration, or distributed system architectures.

62

CI 38–87 · exposure 58 · augmentation 88 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Tech, finance, and professional services sectors are rapidly integrating AI into architecture workflows (prompt-based design, copilot adoption); early-stage production use is visible, though wholesale replacement remains limited by human oversight norms.
Sector adoption velocityclaude-sonnet-53/5Software engineering is a fast-adopting sector for AI coding tools generally, but architecture-level design work lags behind lower-level coding task adoption.
Augmentation potentialclaude-haiku-4-5-202510015/5AI dramatically enhances architect productivity by generating drafts, exploring alternatives, and validating designs in real-time while architects retain judgment; this is among the strongest augmentation use cases in technical work.
Augmentation potentialclaude-sonnet-54/5AI tools meaningfully assist architects by generating design options, documentation, diagrams, and identifying tradeoffs, significantly speeding up exploration and iteration.
Task automatabilityclaude-haiku-4-5-202510015/5AI systems like Claude, GPT-4, and specialized tools can now generate comprehensive system architectures, design diagrams, and architectural documents with minimal human intervention, often meeting or exceeding the 50% time-saving bar for the full design phase including documentation and initial validation.
Task automatabilityclaude-sonnet-52/5High-level architecture design requires synthesizing business constraints, tradeoffs, and organizational context that current AI cannot reliably originate end-to-end, though it can draft components of designs.'
Adoption barriersclaude-haiku-4-5-202510012/5Organizational and liability friction exist (sign-off requirements, risk aversion in critical systems), but no legal licensing barrier mandates human architects; many organizations readily adopt AI-assisted design, especially in less safety-critical domains.
Adoption barriersclaude-sonnet-52/5No licensing requirement, but high liability/error costs for flawed architecture decisions and strong organizational reliance on experienced human architects create moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510015/5AI inference and integration costs for architecture generation are orders of magnitude cheaper than the loaded cost of senior engineers ($150k–$250k+ annually), even accounting for oversight—a single API call costs pennies versus weeks of human work.
Cost vs. human wageclaude-sonnet-52/5Given the need for extensive human review and iteration to validate architectural decisions, the effective cost savings are modest despite cheap raw inference costs.
Technical feasibility todayclaude-haiku-4-5-202510014/5Multiple deployed products (GitHub Copilot, Claude, specialized architectural AI tools) reliably assist in architecture generation and review; however, mature end-to-end production systems handling complex, novel architectures without human review remain limited to narrow domains.
Technical feasibility todayclaude-sonnet-52/5AI coding/design assistants exist and can suggest architecture patterns or generate diagrams, but no product autonomously produces validated production-grade system architectures without heavy senior engineer oversight.

Complete models and simulations, using manual or automated tools, to analyze or predict system performance under different operating conditions.

61

CI 46–75 · exposure 58 · augmentation 88 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Technology and finance sectors are actively deploying AI-assisted simulation and modeling tools in production; cloud platforms and AutoML adoption are accelerating. Engineering teams increasingly use automated hyperparameter search and result analysis, reflecting fast, measurable adoption in digitized enterprise settings.
Sector adoption velocityclaude-sonnet-53/5Software/systems engineering is a fast-adopting sector for AI coding and analysis tools, though specialized simulation and architecture work sees more measured, pilot-stage integration.
Augmentation potentialclaude-haiku-4-5-202510015/5AI significantly augments systems engineers by automating tedious simulation runs, sensitivity analysis, and parameter sweeping while leaving engineers to focus on high-level design, validation, and interpretation. This is a textbook case of productive human-AI collaboration where the human stays in control.
Augmentation potentialclaude-sonnet-54/5AI tools substantially help engineers draft simulation code, explore parameter spaces, and interpret results faster, meaningfully boosting productivity while the engineer retains judgment over model validity.
Task automatabilityclaude-haiku-4-5-202510014/5AI can automate large portions of model setup, parameter tuning, simulation execution, and performance data analysis using tools like AutoML, physics-informed neural networks, and automated hyperparameter optimization. However, human judgment remains valuable for defining problem scope, validating assumptions, and interpreting results in context, preventing a full end-to-end 5 rating.
Task automatabilityclaude-sonnet-53/5AI can generate simulation code, configure performance models, and analyze results, but complex system architecture modeling still requires significant human setup, validation, and domain expertise to ensure accuracy.
Adoption barriersclaude-haiku-4-5-202510012/5Few legal or licensing barriers exist for automating simulations and model building. Organizational friction (preference for human validation, internal process requirements) and error-cost sensitivity in mission-critical systems provide modest friction, but nothing prevents substitution in many contexts.
Adoption barriersclaude-sonnet-52/5No licensing requirement typically applies, but organizational risk tolerance and the need for engineer validation before decisions are made based on simulations create moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510014/5Cloud-based simulation and AutoML tools cost significantly less per analysis than hiring a systems engineer for model development and execution. GPUs and inference are cheap relative to skilled labor, putting AI at a substantial cost advantage, though integration and validation overhead prevents an extreme ratio.
Cost vs. human wageclaude-sonnet-52/5AI can reduce some scripting and analysis time, but the human oversight, domain validation, and iterative tuning required for credible system performance predictions keep costs relatively comparable to skilled engineer time.
Technical feasibility todayclaude-haiku-4-5-202510014/5Multiple mature products (simulation software with AI plugins, AutoML platforms, specialized physics engines with learning components) reliably perform sub-tasks like parameter optimization and result analysis in production. Full end-to-end automation of complex system modeling still encounters edge cases and domain-specific validation requirements that limit production reliability to a 4.
Technical feasibility todayclaude-sonnet-52/5Products like AI coding assistants and simulation tools exist and can accelerate model-building, but no deployed system autonomously completes end-to-end performance modeling for complex systems reliably in production.

Perform security analyses of developed or packaged software components.

57

CI 53–61 · exposure 55 · augmentation 100 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Information, finance, and large enterprise sectors have rapidly adopted automated security scanning tools and CI/CD-integrated SAST/SCA solutions over the past five years. Adoption is production-deep in these high-value sectors, though smaller organizations lag, placing adoption velocity at the higher end of the scale.
Sector adoption velocityclaude-sonnet-54/5Software engineering and security are high-digitization sectors with fast adoption of AI-powered security tooling embedded in CI/CD pipelines, though full replacement of security architects remains limited.
Augmentation potentialclaude-haiku-4-5-202510015/5Security engineers routinely rely on AI-powered tools to identify vulnerabilities, prioritize issues, and suggest fixes, substantially amplifying their ability to analyze large codebases and catch issues earlier. Modern security platforms transform engineer productivity while keeping humans responsible for judgment and architectural decisions.
Augmentation potentialclaude-sonnet-55/5AI significantly augments security analysis by automating vulnerability scanning, flagging suspicious patterns, and suggesting remediations, dramatically increasing the speed and coverage of human security engineers while they retain judgment over prioritization and validation.
Task automatabilityclaude-haiku-4-5-202510013/5AI can assist with portions of security analysis—such as static code analysis, vulnerability scanning, and identifying common patterns—but requires human expertise to contextualize findings, assess business risk, and make recommendations. Full automation meeting the 50% time-saving bar is achievable for routine checks, but comprehensive security analysis typically demands human judgment on architectural implications and threat modeling.
Task automatabilityclaude-sonnet-53/5AI tools can automate significant portions of static/dynamic analysis and vulnerability scanning, but comprehensive security analysis of complex systems still requires human judgment for architecture-level threats, business logic flaws, and novel attack vectors.atable.dev
Adoption barriersclaude-haiku-4-5-202510013/5Regulatory standards (PCI-DSS, SOC 2, HIPAA) often mandate documented security review procedures and may require sign-off by qualified personnel, creating friction. However, no law requires a human to *perform* the analysis itself—only to validate and take responsibility for the outcome, allowing substantial automation.
Adoption barriersclaude-sonnet-53/5No licensing requirement for this specific task, but liability concerns around missed vulnerabilities and compliance requirements (SOC2, FedRAMP, etc.) create organizational friction requiring human sign-off on security assessments.
Cost vs. human wageclaude-haiku-4-5-202510014/5Automated security scanning tools cost significantly less per analysis cycle than hiring security engineers, and integrated platforms reduce the per-check marginal cost substantially. While oversight by humans adds cost, the ratio favors automation for routine scanning tasks, though complex architectural reviews may remain human-cost-competitive.
Cost vs. human wageclaude-sonnet-53/5AI-assisted scanning tools reduce time spent on routine vulnerability detection substantially, but require human security engineers for triage, validation, and deeper analysis, keeping overall costs roughly comparable when factoring in necessary oversight.
Technical feasibility todayclaude-haiku-4-5-202510014/5Mature SAST (static application security testing) and SCA (software composition analysis) tools are deployed in production at scale across enterprises. However, these tools flag issues requiring human interpretation; no product performs end-to-end security analysis and decision-making autonomously, limiting feasibility to 4 rather than 5.
Technical feasibility todayclaude-sonnet-53/5Deployed SAST/DAST tools with AI-enhanced capabilities (e.g., GitHub Copilot security scanning, Snyk, Checkmarx AI features) exist and are used in production, but still generate significant false positives/negatives and miss context-dependent vulnerabilities.

Perform ongoing hardware and software maintenance operations, including installing or upgrading hardware or software.

55

CI 32–77 · exposure 50 · augmentation 88 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510015/5Large tech, finance, and cloud-native enterprises have deep, mature adoption of automated maintenance (continuous deployment, auto-patching, infrastructure-as-code pipelines) spanning years. Smaller organizations and non-tech sectors lag, but in information and professional services, AI-assisted and fully automated maintenance is the standard operational pattern.
Sector adoption velocityclaude-sonnet-53/5IT operations has moderate automation adoption (DevOps, IaC, patch automation) but full autonomous maintenance remains uncommon; pilots of AIOps are growing but not yet dominant.
Augmentation potentialclaude-haiku-4-5-202510015/5AI dramatically enhances human systems engineers via real-time monitoring, predictive alerting, automated root-cause analysis, and guided remediation workflows. Engineers using these tools can manage far larger estates and respond faster; augmentation is transformative and near-universal in tech sectors.
Augmentation potentialclaude-sonnet-54/5AI-assisted scripting, automated patch scheduling, and diagnostic tools significantly speed up planning and troubleshooting even though humans still execute and verify changes.
Task automatabilityclaude-haiku-4-5-202510014/5Current AI agents and orchestration tools can automate large portions of routine maintenance: patching, software upgrades, configuration management, and diagnostics are highly procedural. However, hardware installation often requires physical manipulation and complex troubleshooting in novel environments, preventing a fully autonomous end-to-end solution, though ~50% time savings is achievable through AI-driven patching and deployment automation.
Task automatabilityclaude-sonnet-52/5Physical hardware installation and much software deployment require hands-on execution or environment-specific judgment that current AI cannot fully perform end-to-end, though scripting/automation tools handle parts.'
Adoption barriersclaude-haiku-4-5-202510012/5Few legal or regulatory barriers prevent automation of maintenance tasks. Primary friction is organizational (change management, risk tolerance, preference for human sign-off on critical systems) and technical (legacy systems integration), but these are not hard legal bars and are diminishing as enterprises adopt GitOps and automated deployment.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but organizational risk aversion around production system changes, change-control processes, and physical access needs create moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510014/5Automated patching, monitoring, and software deployment cost a fraction of a systems engineer's loaded wage per task cycle. Once infrastructure-as-code and monitoring systems are in place, incremental AI-driven maintenance is very cheap relative to manual intervention; only complex physical hardware work retains higher labor cost.
Cost vs. human wageclaude-sonnet-52/5Automation tooling reduces some labor cost, but hardware installation and complex troubleshooting still require paid technician time, keeping overall cost comparable to human labor rather than dramatically cheaper.
Technical feasibility todayclaude-haiku-4-5-202510014/5Mature products (Ansible, Puppet, Chef, cloud auto-scaling, and AI-assisted monitoring platforms) reliably perform software updates, patches, and configuration management at scale in production. Hardware diagnostics and remediation are increasingly automated, though complex physical hardware installation still requires human intervention; overall the majority of this task class is deployed and reliable.
Technical feasibility todayclaude-sonnet-52/5Configuration management and patch automation products exist (Ansible, SCCM) but they require human-designed playbooks and oversight; no product autonomously performs full maintenance lifecycles reliably today.

Identify system data, hardware, or software components required to meet user needs.

54

CI 38–70 · exposure 45 · augmentation 88 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Technology-forward sectors (finance, cloud providers, digital-native enterprises) are actively deploying AI-assisted architecture and infrastructure recommendation tools in production environments, showing measurable adoption momentum and productivity gains.
Sector adoption velocityclaude-sonnet-53/5IT and engineering sectors are adopting AI copilots for requirements analysis and system design at a moderate pace, with pilots common but full production reliance still limited.
Augmentation potentialclaude-haiku-4-5-202510015/5AI excels at augmenting system architects by rapidly synthesizing solution architectures, cross-referencing compatibility matrices, and generating candidate component lists from requirements, allowing engineers to focus on design trade-offs and validation while productivity increases significantly.
Augmentation potentialclaude-sonnet-54/5AI tools can significantly speed up requirements documentation, gap analysis, and component research, meaningfully boosting engineer productivity while humans retain final judgment.
Task automatabilityclaude-haiku-4-5-202510014/5Current AI systems (LLMs, agents) can identify required components from detailed requirements specifications by analyzing technical documentation, cross-referencing standards, and inferring architecture patterns with significant automation potential. However, some complex scenarios involving novel constraints or integration edge cases still benefit from expert review, falling short of a full 5.
Task automatabilityclaude-sonnet-52/5This requires eliciting ambiguous stakeholder needs, understanding organizational constraints, and making judgment calls that current AI cannot fully replace end-to-end, though it can assist with parts of the analysis.'},'feasibility':{
Adoption barriersclaude-haiku-4-5-202510012/5While organizational adoption and risk assessment practices create some friction, there are no hard legal or licensure barriers preventing AI from recommending or assisting with system component identification. Organizational preference for human sign-off is common but not mandatory.
Adoption barriersclaude-sonnet-52/5No licensing requirement, but organizational trust, accountability for architecture decisions, and stakeholder relationship management create moderate friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI analysis of component requirements costs orders of magnitude less than hiring senior engineers for routine assessments; inference and cloud consultation APIs are cheap at scale. Full savings are partially offset by integration effort and human verification overhead.
Cost vs. human wageclaude-sonnet-52/5Human engineers still need to conduct interviews, negotiate trade-offs, and validate assumptions, so AI reduces some effort but doesn't yet displace the bulk of billable engineering time.
Technical feasibility todayclaude-haiku-4-5-202510013/5AI-powered tools exist and are deployed (e.g., cloud architecture recommendation engines, AI-assisted infrastructure planning platforms), but they operate with material limitations in understanding organization-specific legacy systems and complex interdependencies, requiring substantial human validation rather than fully autonomous recommendation.
Technical feasibility todayclaude-sonnet-52/5No deployed product autonomously performs full requirements gathering and component identification reliably; existing tools assist analysts but don't replace the human-led discovery process.

Provide advice on project costs, design concepts, or design changes.

49

CI 32–66 · exposure 42 · augmentation 88 · importance 3.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Tech and financial sectors are rapidly adopting AI code assistants and architecture advisors; many DevOps and engineering teams have integrated copilot-style tools into workflows. Adoption is faster and deeper in digitally mature organizations, consistent with fast-mover patterns.
Sector adoption velocityclaude-sonnet-53/5Software/IT engineering is a fast-adopting sector for AI coding tools, but advisory tasks involving cost and design judgment remain in pilot/augmentation stage rather than deep production automation.
Augmentation potentialclaude-haiku-4-5-202510015/5AI excels at augmenting human engineers by rapidly generating candidate designs, cost breakdowns, and alternative architectures for human review and refinement. This is one of the most mature augmentation use cases in software engineering today.
Augmentation potentialclaude-sonnet-54/5LLMs are widely used to draft cost estimates, compare design alternatives, summarize tradeoffs, and prepare talking points, meaningfully speeding up the architect's advisory work while the human retains final judgment.
Task automatabilityclaude-haiku-4-5-202510013/5AI systems can draft cost estimates, outline design concepts, and suggest architectural changes based on technical data and patterns learned from training. However, meaningful advice on cost–benefit tradeoffs, risk assessment, and alignment with organizational constraints typically requires human judgment, limiting autonomous performance below the 50% time-savings bar.
Task automatabilityclaude-sonnet-52/5This requires synthesizing organizational context, stakeholder priorities, and technical tradeoffs into judgment-based advice, which current AI can support but not reliably replace end-to-end.dis Cost estimation and design advice depend heavily on tacit institutional knowledge AI lacks access to.
Adoption barriersclaude-haiku-4-5-202510012/5There is minimal regulatory mandate or licensing requirement for design advice itself; organizational friction around trusting AI recommendations exists but is cultural rather than legal. Most firms can experiment with AI-assisted advisory without legal or compliance barriers.
Adoption barriersclaude-sonnet-53/5No licensing requirement typically, but organizational accountability, liability for costly design errors, and stakeholder trust in a named human architect create real friction against full substitution.
Cost vs. human wageclaude-haiku-4-5-202510014/5Inference cost for AI-generated design sketches and cost advice is very low (fractions of a dollar), whereas a senior systems engineer reviewing the same ground might charge $150–400/hour. Even accounting for oversight, AI scales dramatically cheaper per unit advisory task.
Cost vs. human wageclaude-sonnet-52/5Generating draft advice is cheap, but the necessary human validation, contextual grounding, and accountability for costly engineering decisions keep effective cost comparable to or only modestly below human expert cost.
Technical feasibility todayclaude-haiku-4-5-202510014/5Multiple deployed products (GitHub Copilot, Claude, ChatGPT) can generate credible technical advice on architecture and design changes; some organizations use AI-assisted tools for preliminary cost estimates. The error rate on complex, context-specific recommendations is non-trivial, but the capability is production-adjacent rather than research-stage.
Technical feasibility todayclaude-sonnet-52/5AI coding/design assistants can suggest architectural patterns or flag cost drivers, but no deployed product autonomously provides authoritative project cost or design-change advice in production engineering workflows.

Design and conduct hardware or software tests.

47

CI 37–57 · exposure 42 · augmentation 88 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Tech and finance sectors are piloting AI-assisted testing (code completion, test generation), but production displacement remains limited. Most organizations still employ human test engineers for architecture and strategy; AI is augmentative rather than substitutive.
Sector adoption velocityclaude-sonnet-54/5Software engineering and QA sectors have rapidly adopted AI coding and testing assistants (Copilot, automated test generators) as part of mainstream developer tooling in the past few years.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly assists test engineers through code completion, test case suggestions, and coverage analysis. Engineers using these tools can author and review test suites faster while maintaining quality and architectural control.
Augmentation potentialclaude-sonnet-55/5AI tools substantially boost productivity in generating test cases, identifying edge cases, and scaffolding test suites, while engineers remain responsible for verifying coverage and correctness.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can generate test cases and scripts, designing comprehensive tests requires deep understanding of system requirements, edge cases, and business logic. Current AI struggles with novel architectural decisions and complex integration testing scenarios that demand human judgment.
Task automatabilityclaude-sonnet-53/5AI can generate test cases, unit tests, and even fuzzing scripts, saving significant time, but designing comprehensive test strategies for complex systems still requires human architectural judgment and domain knowledge.
Adoption barriersclaude-haiku-4-5-202510013/5Test quality directly impacts system reliability and liability; organizations typically require human sign-off on test plans and results. Professional judgment in test strategy and risk assessment creates moderate friction against full automation.
Adoption barriersclaude-sonnet-52/5No licensing requirement blocks AI use in test design, though critical systems (safety, security) impose review and validation requirements that create moderate friction before AI-generated tests are trusted.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI code generation tools reduce some writing effort, but human engineers must still architect tests, validate coverage, and fix false positives. The total cost (tools + human review + oversight) remains comparable to skilled manual test design.
Cost vs. human wageclaude-sonnet-53/5AI-assisted test generation reduces engineer hours, but integration, validation, and oversight of AI-generated tests still require substantial engineering time, keeping costs roughly comparable to fully human-led testing for complex systems.
Technical feasibility todayclaude-haiku-4-5-202510013/5Deployed products (GitHub Copilot, ChatGPT) can assist with test code generation and suggest test structures, but they produce material error rates and miss coverage gaps that human engineers must catch. No product reliably designs end-to-end test suites without human oversight.
Technical feasibility todayclaude-sonnet-53/5Products like Copilot, automated test generation tools, and CI/CD integrated AI testing agents exist and are used in production, but they handle narrow subsets of test design rather than full end-to-end test strategy creation reliably.

Evaluate current or emerging technologies to consider factors such as cost, portability, compatibility, or usability.

47

CI 36–57 · exposure 38 · augmentation 88 · importance 3.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Technology and engineering firms are actively piloting AI-assisted evaluation tools, and major cloud vendors now embed AI-driven recommendation engines in architecture workflows. Adoption is accelerating in digitized enterprise environments, though final decisions remain human-centered.
Sector adoption velocityclaude-sonnet-53/5Software/IT sectors are fast adopters of AI-assisted research and coding tools, though full evaluation workflows are still human-led with AI as a supporting tool.
Augmentation potentialclaude-haiku-4-5-202510014/5AI excels at augmenting this task: summarizing vendor documentation, cross-referencing compatibility matrices, surfacing cost-benefit trade-offs, and flagging emerging options. Human architects remain in control but gain significant productivity gains from AI-assisted data synthesis and preliminary analysis.
Augmentation potentialclaude-sonnet-55/5AI is highly effective at rapidly surfacing technology comparisons, cost data, compatibility notes, and usability considerations, significantly speeding up the engineer's research process.
Task automatabilityclaude-haiku-4-5-202510013/5AI can automate parts of technology evaluation—gathering specifications, comparing costs, and producing compatibility matrices—but requires human judgment on strategic trade-offs, organizational context, and emerging tech viability. The task has both automatable data-synthesis components and judgment-heavy decisions that limit end-to-end automation below the 50% threshold.
Task automatabilityclaude-sonnet-52/5AI can research and summarize technology comparisons, but the actual evaluation involves synthesizing organizational context, stakeholder priorities, and judgment calls that current systems cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510013/5Technology decisions often require institutional sign-off and accountability, and risk-averse organizations prefer human architects to own critical infrastructure choices. Regulatory and liability concerns exist but are typically resolved with human review rather than strict legal barriers to AI use.
Adoption barriersclaude-sonnet-52/5No licensing requirement exists, but organizational trust and accountability for architecture decisions create some friction against fully automating this judgment task.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI-assisted evaluation (literature synthesis, spec comparison, initial analysis) is comparable in cost to junior engineer time for the same depth of work, but senior expertise and final sign-off remain human-intensive. The cost advantage is modest and highly dependent on task scope.
Cost vs. human wageclaude-sonnet-53/5AI can cheaply generate comparison matrices and summaries, but human oversight and validation of technical tradeoffs still add substantial cost, making the net savings moderate rather than dramatic.
Technical feasibility todayclaude-haiku-4-5-202510013/5Deployed products (LLMs with web search, enterprise software-selection tools) can assist with data gathering and preliminary comparisons, but real-world technology evaluation involves nuanced domain expertise and context-specific requirements that current systems handle inconsistently. Narrow scope and frequent manual intervention are common.
Technical feasibility todayclaude-sonnet-52/5Deployed AI research/analysis tools can gather comparative data on technologies, but no product reliably performs holistic technology evaluation with sound engineering judgment in production settings.

Communicate project information through presentations, technical reports, or white papers.

45

CI 32–57 · exposure 42 · augmentation 88 · importance 3.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Many tech companies are piloting AI-assisted drafting and presentation tools, but adoption remains cautious; most production use is human-in-the-loop assistance rather than autonomous generation, reflecting the high stakes of technical communication.
Sector adoption velocityclaude-sonnet-54/5Software/IT and engineering firms are fast adopters of AI writing and presentation tools, with widespread use of tools like Copilot, ChatGPT, and Gamma in professional documentation workflows.
Augmentation potentialclaude-haiku-4-5-202510014/5AI excels at accelerating first drafts, outlining structure, generating visual descriptions, and catching prose issues, making it a strong productivity multiplier when engineers retain control over technical accuracy and message framing.
Augmentation potentialclaude-sonnet-55/5AI substantially speeds up drafting, formatting, and summarizing technical content into presentations and reports, with the engineer remaining responsible for accuracy and final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can draft reports and generate presentation outlines from structured data, this task fundamentally requires human judgment about audience, messaging priorities, and technical accuracy verification. Current systems cannot reliably produce polished, coherent deliverables end-to-end that meet the 50% time-saving threshold without substantial human revision and validation.
Task automatabilityclaude-sonnet-53/5AI can draft presentations, reports, and white papers from source material and bullet points, but requires human input of accurate technical content and review for correctness, limiting full end-to-end automation.
Adoption barriersclaude-haiku-4-5-202510014/5Technical presentations and white papers often carry organizational credibility and liability stakes; signing off on AI-generated claims requires expert review, and organizational culture values authorship and accountability tied to real engineers, creating friction against full automation.
Adoption barriersclaude-sonnet-52/5No licensing requirement blocks AI-assisted communication, though organizational sign-off and accountability for technical accuracy create some friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5While API costs are low, the overhead of prompt engineering, fact-checking, revision cycles, and domain expert oversight often approaches or exceeds the cost of a domain expert drafting directly, especially for high-stakes technical communication.
Cost vs. human wageclaude-sonnet-53/5AI drafting tools are cheap per document, but the need for expert review, fact-checking, and iteration on technical content keeps overall cost roughly comparable to a partially-augmented human workflow.
Technical feasibility todayclaude-haiku-4-5-202510013/5AI writing assistants (ChatGPT, Claude, Copilot) are deployed and used for drafting technical content, but material error rates, inconsistencies in technical accuracy, and the need for human oversight of complex claims limit their reliability for production-quality white papers and critical presentations.
Technical feasibility todayclaude-sonnet-53/5Deployed tools (LLM writing assistants, slide generators) reliably produce drafts of reports and presentations today, but engineers still must supply accurate technical details and heavily edit for domain-specific accuracy.

Develop efficient and effective system controllers.

44

CI 30–59 · exposure 45 · augmentation 75 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Technology firms and digital-forward companies are increasingly using AI-assisted code generation and simulation for controller development, but adoption remains uneven across industries; financial services and embedded systems sectors show faster uptake than traditional manufacturing.
Sector adoption velocityclaude-sonnet-52/5Systems engineering and embedded/controller design remain in early-stage AI adoption compared to software development broadly, with pilots more common than production-scale AI-driven design.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly augments controller development by accelerating code prototyping, running large-scale simulations, and suggesting optimizations based on performance data, while the engineer retains critical judgment on architecture and safety requirements.
Augmentation potentialclaude-sonnet-54/5AI tools substantially assist with code generation, simulation, debugging, and documentation for controller development, meaningfully boosting engineer productivity while humans retain design responsibility.
Task automatabilityclaude-haiku-4-5-202510014/5A significant portion of controller development—such as code generation, optimization of standard controller types, simulation testing, and documentation—can be automated with AI systems today, though human oversight of domain-specific requirements and safety-critical tuning often remains necessary, meeting near-full automation thresholds.
Task automatabilityclaude-sonnet-52/5Developing system controllers involves deep domain-specific engineering judgment, hardware/software integration, and iterative testing that current AI can assist but not fully replace end-to-end at equal quality with major time savings.
Adoption barriersclaude-haiku-4-5-202510013/5Many controller applications (aerospace, automotive, industrial) carry regulatory requirements and safety certification demands that mandate human sign-off and liability accountability, creating friction; however, most controller work sits in less regulated contexts where adoption is easier.
Adoption barriersclaude-sonnet-53/5While not formally licensed in most cases, system controllers often affect safety-critical infrastructure, requiring rigorous validation and accountability that create organizational and liability friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI-assisted controller development (code generation, simulation, and optimization) approaches cost parity with experienced engineer time when integration and validation overhead are included, though it varies by controller complexity.
Cost vs. human wageclaude-sonnet-52/5Given the need for extensive human oversight, testing, and domain expertise to validate controller correctness and safety, AI cost savings are modest relative to skilled engineer wages.
Technical feasibility todayclaude-haiku-4-5-202510013/5Deployed tools exist for code generation, testing automation, and performance optimization in controller systems, but they typically require substantial human expertise to guide domain choices and validate safety properties; production use is growing but not yet universal across complex control scenarios.
Technical feasibility todayclaude-sonnet-52/5AI coding assistants can help draft control logic or firmware snippets, but no deployed product autonomously designs, validates, and deploys full system controllers reliably in production.

Provide technical guidance or support for the development or troubleshooting of systems.

42

CI 36–49 · exposure 30 · augmentation 88 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Tech companies actively integrate AI-assisted development tools into workflows, but deployment of fully autonomous system guidance remains limited; most adoption is augmentative (co-pilots) rather than replacive, reflecting the maturity and risk profile of the sector.
Sector adoption velocityclaude-sonnet-54/5IT and software engineering sectors are fast adopters of AI coding and troubleshooting assistants, with widespread pilot and production use of tools like Copilot and AI-driven diagnostics.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly augments engineers through rapid code analysis, documentation generation, troubleshooting suggestions, and architectural pattern matching, materially raising productivity while engineers retain decision-making authority over critical guidance.
Augmentation potentialclaude-sonnet-55/5AI significantly augments this task today through code analysis, log parsing, error diagnosis suggestions, and knowledge retrieval, substantially speeding up troubleshooting while the engineer remains responsible for final decisions.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with troubleshooting scripts and documentation, genuine technical guidance requires understanding organizational context, legacy system constraints, and nuanced architectural decisions that vary widely across deployments. Current AI systems cannot reliably handle end-to-end system troubleshooting with the 50% time-saving threshold without human verification of critical decisions.
Task automatabilityclaude-sonnet-52/5Providing technical guidance and troubleshooting complex systems requires deep contextual understanding, judgment, and interaction with unique organizational infrastructure that current AI cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510013/5While no formal licensing requirement mandates human engineers perform this task, organizational risk tolerance, liability concerns around system failures, and the established expectation that senior engineers provide guidance create meaningful friction against full automation.
Adoption barriersclaude-sonnet-52/5No licensing requirement exists for this work, but organizational trust, liability for system failures, and the need for accountable expertise create moderate friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI inference and integration costs for continuous technical guidance are becoming comparable to junior engineer wages, but the overhead of human oversight, validation, and correction keeps the effective cost ratio near parity rather than substantially cheaper.
Cost vs. human wageclaude-sonnet-53/5AI tools can reduce time spent on some diagnostic and research tasks, but the human engineer's oversight, validation, and integration work still dominate cost, keeping the ratio only moderately favorable.
Technical feasibility todayclaude-haiku-4-5-202510013/5Products like GitHub Copilot and ChatGPT demonstrate moderate capability in code debugging and technical explanation, but they operate with notable error rates on complex architectural questions and often require expert validation before implementation in production systems.
Technical feasibility todayclaude-sonnet-52/5AI coding assistants and diagnostic tools exist and are used for narrow troubleshooting help, but no deployed product reliably provides comprehensive technical guidance across diverse, complex systems without significant human oversight.

Develop application-specific software.

42

CI 28–57 · exposure 38 · augmentation 88 · importance 3.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Technology sectors are rapidly piloting and integrating code-generation copilots into development workflows, but actual displacement of architects and engineers remains limited. Adoption is strong in code-completion and documentation tasks; architectural and integration decisions are still human-led.
Sector adoption velocityclaude-sonnet-54/5Software engineering is a fast-adopting, highly digitized field with widespread production use of AI coding assistants across tech companies and beyond.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly augments developer productivity through code completion, test generation, documentation drafting, and boilerplate creation, allowing engineers to focus on design and integration logic. LLMs have demonstrably transformed day-to-day coding efficiency while humans retain architectural and validation responsibilities.
Augmentation potentialclaude-sonnet-55/5AI coding assistants dramatically speed up boilerplate generation, debugging, and prototyping for software engineers while the engineer retains control over architecture and correctness.
Task automatabilityclaude-haiku-4-5-202510012/5Current AI can assist with code generation and some boilerplate creation, but end-to-end application software development requires architectural decisions, complex debugging, integration testing, and validation that AI cannot reliably complete without substantial human oversight and iteration. The domain-specific logic and cross-system concerns remain firmly in human territory.
Task automatabilityclaude-sonnet-53/5AI coding assistants can generate significant portions of application-specific code, but full end-to-end development including requirements analysis, architecture, integration, and debugging still requires substantial human effort for anything beyond simple applications.tags
Adoption barriersclaude-haiku-4-5-202510014/5Liability and error-cost asymmetry are severe: shipped applications directly impact business operations and user safety, requiring human sign-off and professional accountability that cannot be delegated to AI. Regulatory requirements (security, compliance, data handling) and organizational risk tolerance create strong friction against full automation.
Adoption barriersclaude-sonnet-52/5No licensing requirement to write software, though some regulated industries impose validation and liability requirements on custom software; overall barriers are low to moderate.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI code-generation tools reduce time on specific coding segments, but integration, testing, architecture review, and debugging still require expensive senior engineers. The all-in cost of AI-assisted development (tool licensing, human oversight, rework) remains comparable to or sometimes exceeds hiring experienced engineers directly.
Cost vs. human wageclaude-sonnet-53/5AI coding tools are relatively cheap per token but require paid licenses plus significant engineer time for review, correction, and integration, making net cost savings moderate rather than order-of-magnitude for full application development.
Technical feasibility todayclaude-haiku-4-5-202510012/5Code generation tools (GitHub Copilot, Claude, ChatGPT) exist and are deployed in production, but they function as assistants for fragments rather than autonomous application builders. Reliable end-to-end application delivery without human architects and engineers directing requirements, design, and validation remains absent from production systems.
Technical feasibility todayclaude-sonnet-53/5Products like GitHub Copilot, Cursor, and Claude Code are deployed widely in production and reliably assist with code generation, but they don't autonomously develop complete application-specific software without heavy human oversight and iteration.

Verify stability, interoperability, portability, security, or scalability of system architecture.

39

CI 28–50 · exposure 38 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Tech companies and large enterprises actively adopt automated verification tools in CI/CD pipelines, but widespread production deployment of fully autonomous architectural validation remains limited. Most organizations still require senior engineer review and approval before deployment.
Sector adoption velocityclaude-sonnet-53/5Software engineering and DevOps sectors show moderate-to-fast adoption of AI-assisted code review and testing tools, though full architectural verification workflows remain largely human-driven with AI as a pilot-stage addition.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly augments architects through automated testing, vulnerability scanning, performance modeling, and interoperability checking, allowing engineers to focus on complex design decisions and trade-offs while the tool handles routine verification.
Augmentation potentialclaude-sonnet-54/5AI tools significantly speed up static analysis, vulnerability scanning, dependency checking, and generating test scenarios, meaningfully boosting an architect's ability to verify system qualities while human judgment remains central.
Task automatabilityclaude-haiku-4-5-202510013/5Current AI can automate significant portions of stability and security verification through static analysis, code scanning, and automated testing frameworks, achieving partial time savings. However, complex architectural trade-offs, novel interoperability scenarios, and scalability validation under real-world conditions still require substantial human judgment and oversight.
Task automatabilityclaude-sonnet-52/5AI can assist with automated testing, static analysis, and generating verification scripts, but comprehensive architectural verification requires holistic judgment across quality attributes that current tools cannot fully replace.
Adoption barriersclaude-haiku-4-5-202510014/5System architecture verification typically requires licensed or certified engineers to sign off on critical production systems, especially in regulated industries (finance, healthcare, aerospace). Organizational liability, risk aversion, and compliance requirements create substantial friction against full automation.
Adoption barriersclaude-sonnet-53/5No formal licensing requirement, but liability for security/scalability failures in production systems creates strong organizational incentive for experienced human oversight and sign-off.
Cost vs. human wageclaude-haiku-4-5-202510013/5Automated verification tools are relatively cheap to operate at scale, but integration with existing systems, configuration tuning, and human review of findings approach human wage parity for the full end-to-end verification task.
Cost vs. human wageclaude-sonnet-52/5AI-assisted scanning tools reduce some manual effort but still require significant human architect time for interpretation, tradeoff analysis, and sign-off, keeping costs comparable to or only modestly below human-only effort.
Technical feasibility todayclaude-haiku-4-5-202510013/5Mature products exist (e.g., static analysis tools, automated security scanning, load-testing platforms) that perform parts of this task in production. However, they operate within narrow technical scopes and still require human architects to interpret results, validate findings, and make final architectural decisions.
Technical feasibility todayclaude-sonnet-52/5Deployed tools (static analyzers, security scanners, load testing platforms) address individual sub-checks like security or scalability, but no integrated product reliably verifies overall system architecture across all these dimensions simultaneously.

Provide customers or installation teams guidelines for implementing secure systems.

37

CI 25–50 · exposure 38 · augmentation 63 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Security engineering remains a high-touch, expert-driven domain with slow AI adoption. While some firms pilot AI-assisted documentation, most still rely on human architects to author guidelines. Regulatory and liability concerns keep adoption primarily in support roles rather than autonomous generation.
Sector adoption velocityclaude-sonnet-53/5IT/security consulting and enterprise architecture sectors are adopting AI tools for documentation and knowledge assistance at a moderate pace, with pilots more common than full production reliance.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist engineers by drafting boilerplate sections, generating compliance checklists, and suggesting threat scenarios, which accelerates guideline authoring. However, the human engineer must validate, customize, and verify security correctness, making this assistive rather than transformative.
Augmentation potentialclaude-sonnet-54/5AI substantially speeds up drafting, researching best practices, and structuring guidelines, letting the architect focus on customization and validation.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can generate security best practices and documentation templates, providing contextually appropriate guidelines for diverse customer environments requires understanding specific threat models, compliance requirements, and existing infrastructure. Current AI systems struggle with the necessary customization and fail-safe verification that production security guidance demands.
Task automatabilityclaude-sonnet-53/5AI can draft generic secure-implementation guidelines and checklists quickly, but tailoring to specific customer infrastructure, compliance context, and edge cases still requires expert judgment and validation.'
Adoption barriersclaude-haiku-4-5-202510014/5Security implementation guidelines carry substantial liability: incorrect guidance can compromise customer infrastructure and data. Customers and regulators expect human expert accountability, and many organizations legally require a licensed engineer to author and stand behind security policies, creating strong adoption friction.
Adoption barriersclaude-sonnet-53/5No licensing mandate universally requires a human to author these guidelines, but liability for security failures and customer trust in expert judgment create meaningful friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5While AI drafting tools have low inference cost, the final security guidelines require expert review, testing, and validation. The overhead of verification and the liability cost if AI-generated guidance fails means the all-in cost remains comparable to or exceeds having engineers draft guidelines directly.
Cost vs. human wageclaude-sonnet-53/5AI can cheaply generate draft guidelines, but the human architect's review, customization, and liability oversight remain a significant cost component, making overall savings moderate rather than dramatic.
Technical feasibility todayclaude-haiku-4-5-202510012/5AI writing assistants exist but are not deployed to autonomously produce security guidelines in production. Security guidelines require domain expertise, regulatory alignment, and legal accountability; organizations still rely on human experts to author and sign off on implementation guidance, not AI-generated documents.
Technical feasibility todayclaude-sonnet-53/5AI coding/security assistants and documentation generators are used in production to draft security guidance, but reliable, context-accurate guidance for complex enterprise systems still needs human architect review.

Train system users in system operation or maintenance.

35

CI 29–41 · exposure 25 · augmentation 75 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While training augmentation (AI-assisted content creation, chatbots) is spreading in information-sector organizations, actual replacement of human trainers in production remains rare. Most sectors still rely on live instructors, and adoption of autonomous training systems lags behind other automation patterns.
Sector adoption velocityclaude-sonnet-53/5IT and professional services sectors are moderately fast adopters of AI for documentation and support, but live training delivery still commonly involves human trainers or blended approaches.chunk
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly assists human trainers by generating course materials, creating interactive simulations, answering routine questions via chatbots, and personalizing content recommendations. These tools meaningfully amplify trainer productivity while humans retain responsibility for live instruction, assessment, and learner support.
Augmentation potentialclaude-sonnet-54/5AI can significantly assist by generating training manuals, quizzes, video scripts, and FAQ chatbots, greatly speeding up preparation while a human still delivers or oversees actual training.chunk
Task automatabilityclaude-haiku-4-5-202510012/5Training requires responsive interaction with learners, adaptation to their comprehension level, and addressing their specific questions—capabilities at which current AI performs poorly. While AI can generate static training materials and basic instructional content, it cannot reliably manage the dynamic, two-way dialogue and real-time troubleshooting feedback that effective system training demands, nor achieve the ≥50% time-saving bar for the full task.
Task automatabilityclaude-sonnet-52/5Training involves live delivery, adapting to trainee questions, hands-on demonstration, and reading audience comprehension, which current AI cannot fully replicate end-to-end though it can produce training materials.chunk
Adoption barriersclaude-haiku-4-5-202510014/5Organizations typically require human trainers to maintain accountability, comply with certification standards, and ensure compliance—especially in regulated or safety-critical contexts. User confidence and organizational liability concerns create substantial friction against full automation, even where technically possible.
Adoption barriersclaude-sonnet-52/5No licensing requirement typically governs this task, but organizational preference for human trainers who can adapt live and build trust with users creates moderate friction.chunk
Cost vs. human wageclaude-haiku-4-5-202510013/5AI-generated training materials and asynchronous support can reduce costs modestly compared to one-on-one instructor time, but the requirement for customization, error correction, and hands-on labs means human oversight and supplementation remain necessary, placing overall cost closer to parity than strong advantage.
Cost vs. human wageclaude-sonnet-53/5AI can cheaply generate training content and FAQs, but human-led sessions or oversight are still needed for complex system training, keeping overall cost roughly comparable once integration and review are factored in.chunk
Technical feasibility todayclaude-haiku-4-5-202510012/5No production systems reliably perform live, interactive training of system users end-to-end. Deployed products can auto-generate documentation or provide chatbot FAQs, but these are narrow supplements rather than substitutes for human-led training, which demands contextual adaptability and presence that current systems lack at scale.
Technical feasibility todayclaude-sonnet-52/5AI tools exist for generating documentation, tutorials, and chatbots that answer questions, but no deployed product reliably conducts full interactive training and hands-on troubleshooting sessions for complex systems.chunk

Investigate system component suitability for specified purposes, and make recommendations regarding component use.

33

CI 25–41 · exposure 25 · augmentation 75 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While IT organizations adopt component-selection tools, autonomous or near-autonomous AI recommendation systems remain rare in production; most adoption is in supplementary analysis and filtering rather than replacing architect judgment. Uptake is slow due to liability and expertise requirements.
Sector adoption velocityclaude-sonnet-53/5IT and engineering sectors show moderate AI adoption for research and documentation tasks, though architectural decision-making remains largely human-led with pilots ongoing.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can significantly augment architects by rapidly comparing component specifications, flagging compatibility issues, analyzing benchmarks, and surfacing options that meet performance criteria. An architect working with such tools can evaluate vastly more component combinations and trade-offs than manual research alone.
Augmentation potentialclaude-sonnet-54/5AI tools significantly speed up literature/spec review, compatibility checks, and drafting comparative analyses, meaningfully boosting engineer productivity while the human retains final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can analyze component specifications and flag compatibility issues, the task requires contextual judgment about business requirements, legacy constraints, and organizational priorities that typically demand human expert review. AI can assist in data gathering and filtering but cannot independently recommend suitable components end-to-end at equal quality without substantial human oversight.
Task automatabilityclaude-sonnet-52/5AI can help research and compare components, but making authoritative recommendations requires integrating organizational constraints, tacit knowledge, and risk judgment that current systems cannot fully replicate end-to-end.'
Adoption barriersclaude-haiku-4-5-202510014/5Significant organizational and liability barriers exist: recommendations directly affect system performance, security, and costs; architects are professionally responsible for their recommendations; and most enterprises require human sign-off on infrastructure decisions for compliance and accountability reasons.
Adoption barriersclaude-sonnet-52/5No strict licensing requirement, but organizational risk-aversion and accountability for architecture decisions create moderate friction against pure AI-driven recommendations.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for component analysis are relatively inexpensive, but the task requires expert architects to validate recommendations, interpret results, and take responsibility for decisions. The combined cost of AI analysis plus human expert review approaches or exceeds the cost of having the engineer do the work directly.
Cost vs. human wageclaude-sonnet-53/5AI can cheaply generate comparative research and draft recommendations, but human validation, testing, and sign-off still add substantial cost, making overall savings moderate rather than dramatic.
Technical feasibility todayclaude-haiku-4-5-202510012/5Products like procurement AI and component database systems exist, but they operate narrowly—matching specs to catalogs—rather than performing the full investigative task of evaluating suitability across competing requirements, risk factors, and organizational fit. No deployed product reliably makes architectural recommendations independently.
Technical feasibility todayclaude-sonnet-52/5Products like coding assistants and technical chatbots can surface component comparisons, but no deployed system reliably performs full suitability investigations and recommendations autonomously in production.

Develop or approve project plans, schedules, or budgets.

32

CI 28–36 · exposure 30 · augmentation 75 · importance 3.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Tech-forward companies are adopting AI planning assistants in pilots and for draft generation, but widespread production use remains limited. Most organizations still require human-led planning with AI as a drafting aid rather than decision-maker, reflecting moderate adoption rather than deep market penetration.
Sector adoption velocityclaude-sonnet-53/5Tech/engineering sectors are moderately fast adopters of AI-assisted planning tools, but full automation of approval authority remains rare and cautious.
Augmentation potentialclaude-haiku-4-5-202510014/5AI excels at augmenting this task by generating baseline schedules, flagging resource conflicts, modeling scenarios, and updating budget forecasts in real time. Engineers and architects can validate and refine AI suggestions much faster than building plans from scratch, substantially raising their productivity.
Augmentation potentialclaude-sonnet-54/5AI can significantly speed up drafting schedules, estimating costs, and flagging risks, meaningfully boosting the human planner's productivity while they retain decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5AI can draft project plans and generate initial schedules using templates and historical data, but approval decisions and budget validation require human judgment, risk assessment, and stakeholder accountability that AI cannot reliably replace. The task involves tradeoff analysis and organizational context that prevents full automation.
Task automatabilityclaude-sonnet-52/5Planning and budgeting require organizational context, stakeholder negotiation, and risk judgment that AI cannot fully replicate; AI can draft schedules but cannot approve them autonomously with equal quality.
Adoption barriersclaude-haiku-4-5-202510014/5Project and budget approval is often a contractual and legal responsibility tied to specific roles; sign-off authority typically rests with licensed architects or senior engineers who bear accountability. Organizational governance and liability frameworks create strong resistance to full delegation to AI.
Adoption barriersclaude-sonnet-54/5Approval of budgets and project plans typically requires accountable, authorized personnel due to liability, contractual, and organizational governance requirements.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI-assisted planning tools cost less than hiring dedicated planning staff, but the human engineer or architect must review, refine, and approve all outputs. The savings are moderate because significant human oversight remains necessary, making total delivered cost roughly comparable to traditional human effort.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply generate draft schedules, but the approval and accountability portion still requires senior engineer time, keeping overall cost comparable to human-driven processes.
Technical feasibility todayclaude-haiku-4-5-202510013/5Generative AI tools and project management software can produce draft plans and schedules; some organizations use AI to suggest resource allocation and identify scheduling conflicts. However, production systems still show gaps in handling complex dependencies, stakeholder constraints, and final approval workflows at scale.
Technical feasibility todayclaude-sonnet-52/5Project management tools with AI features (e.g., schedule generation, resource forecasting) exist but are used as aids, not as reliable autonomous planners/approvers in production.

Evaluate existing systems to determine effectiveness, and suggest changes to meet organizational requirements.

31

CI 25–38 · exposure 25 · augmentation 63 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While IT organizations increasingly use monitoring and diagnostic tools, actual delegation of evaluation and change recommendation remains limited. Most deployments are still at the pilot/assistive stage in professional technology services; deep production automation of this judgment task is not widespread.
Sector adoption velocityclaude-sonnet-53/5IT and software engineering sectors are moderately fast adopters of AI tooling for code review and system analysis, though full evaluation workflows still involve significant human oversight and pilots rather than mature deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully assist by analyzing performance metrics, generating diagnostic reports, and highlighting anomalies that architects review and act on. However, the core judgment of 'effectiveness' and organizational fit requires human expertise; augmentation is meaningful but not transformative.
Augmentation potentialclaude-sonnet-54/5AI tools can significantly speed up system analysis, flag inefficiencies, summarize logs/metrics, and draft improvement proposals, meaningfully boosting engineer productivity while humans retain decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5Evaluating system effectiveness and suggesting changes require nuanced understanding of organizational context, priorities, and trade-offs that exceed current AI capabilities. While AI can analyze logs and identify performance metrics, the judgment-heavy task of determining 'effectiveness' relative to shifting business requirements and recommending strategic changes remains firmly in the human domain.
Task automatabilityclaude-sonnet-52/5AI can help gather metrics and suggest generic improvements, but evaluating a specific organization's system effectiveness against its unique requirements requires deep contextual judgment, stakeholder input, and accountability that current AI cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Systems architects must often sign off on changes and bear responsibility for effectiveness determinations; organizations place high liability on this role. Regulatory requirements, change control boards, and organizational governance typically mandate human accountability for system evaluation and recommendations.
Adoption barriersclaude-sonnet-52/5No licensing requirement, but organizational risk tolerance, need for accountability in architecture decisions, and stakeholder trust create moderate friction against pure AI-driven recommendations.
Cost vs. human wageclaude-haiku-4-5-202510012/5Current AI tools for system analysis (monitoring dashboards, log analysis) still require substantial human oversight and validation. The cost of AI analysis plus required expert review typically exceeds the cost of direct human evaluation, especially given liability exposure from flawed recommendations.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply generate analysis and suggestions, but the human engineer still must validate, contextualize, and integrate findings with organizational strategy, so overall cost savings are moderate rather than dramatic.
Technical feasibility todayclaude-haiku-4-5-202510012/5No production systems reliably perform comprehensive system evaluation and strategic recommendation generation at the organizational level today. Narrow diagnostic tools exist, but end-to-end evaluation of 'effectiveness' with credible change suggestions requires contextual reasoning beyond deployed AI capabilities.
Technical feasibility todayclaude-sonnet-52/5Deployed tools exist for code analysis, architecture review assistance, and monitoring dashboards, but no production system autonomously performs holistic system evaluation and recommends organizationally-tailored changes reliably.

Direct the analysis, development, and operation of complete computer systems.

30

CI 28–32 · exposure 25 · augmentation 75 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Tech and financial sectors have adopted AI for technical analysis and design assistance at medium pace, with pilots and pockets of production use. However, deep displacement in the direction and strategic oversight role itself remains limited, as organizational structures still vest final decision authority in human architects. Adoption is mixed rather than rapid or laggard.
Sector adoption velocityclaude-sonnet-53/5Software/IT sectors show fast AI tool adoption for coding and analysis support, but adoption of AI for the managerial 'directing' aspect of system architecture is still nascent.
Augmentation potentialclaude-haiku-4-5-202510014/5AI augmentation is strong here: systems engineers widely use AI for documentation generation, design pattern suggestions, code reviews, and architecture analysis. These tools genuinely raise productivity in the analytical phases while the human architect retains full direction and accountability, representing meaningful human-in-the-loop productivity gains.
Augmentation potentialclaude-sonnet-54/5AI substantially aids the underlying analysis, documentation, and development work an architect directs, improving productivity even though the direction/oversight remains human-led.
Task automatabilityclaude-haiku-4-5-202510012/5This task requires strategic direction, oversight, and judgment across the full lifecycle of complex systems—from requirements gathering to deployment and ongoing operations. While AI can assist with individual technical analyses and recommendations, directing the complete system (prioritization, trade-offs, accountability) remains fundamentally a human leadership function that current AI cannot replace end-to-end with 50% time savings at equal quality.
Task automatabilityclaude-sonnet-52/5This is a high-level managerial/architectural directing task requiring judgment, stakeholder coordination, and accountability that current AI cannot autonomously perform end-to-end.leaving only sub-components (e.g., drafting specs) automatable.
Adoption barriersclaude-haiku-4-5-202510014/5Significant organizational and liability barriers protect this task: systems engineers/architects bear responsibility and accountability for system failures, data security, and business continuity. Organizations typically require a licensed or credentialed human professional to sign off on critical system decisions, and regulatory/contractual obligations often mandate human expert oversight of complete system direction.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but strong organizational and liability barriers exist since this role involves accountability for enterprise systems and strategic decisions that firms are reluctant to delegate to unsupervised AI.
Cost vs. human wageclaude-haiku-4-5-202510012/5While AI can reduce costs on specific technical subtasks (code generation, documentation), the core function—directing strategy and operations of complete systems—still requires experienced human engineers/architects whose loaded wages far exceed current AI inference costs. AI cannot fully displace this role, keeping overall cost ratio unfavorable for full substitution.
Cost vs. human wageclaude-sonnet-52/5Because AI cannot substitute for the directing function itself, organizations still pay for senior engineers/architects; AI tools only marginally reduce supporting labor costs.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed AI can support subsystems of technical analysis and code generation, but no production system today reliably 'directs' the analysis, development, and operation of complete computer systems independently. Tools exist for components (code review, design suggestions, monitoring), but orchestrating the entire direction and accountability chain remains beyond current deployed capabilities.
Technical feasibility todayclaude-sonnet-52/5AI coding assistants and design tools exist but no deployed product directs or manages complete systems analysis/operation autonomously; humans remain firmly in the directing role.

Direct the installation of operating systems, network or application software, or computer or network hardware.

30

CI 28–32 · exposure 25 · augmentation 63 · importance 4.1/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5IT departments use AI for planning and documentation, but actual installation direction remains largely human-led in production settings. Adoption of AI assistants is moderate; full automation is not yet pursued at scale due to risk and accountability requirements.
Sector adoption velocityclaude-sonnet-53/5IT/software sectors are fast adopters of automation tooling (CI/CD, IaC) but the specific 'directing' oversight role remains human-centric with moderate uptake of AI-assisted orchestration tools.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist engineers by generating installation scripts, predicting conflicts, checking configuration syntax, and providing decision support, raising engineer productivity in directing complex multi-system deployments.
Augmentation potentialclaude-sonnet-54/5AI-driven infrastructure-as-code, deployment scripts, and copilot tools significantly speed up planning, scripting, and troubleshooting for installations, meaningfully boosting engineer productivity while humans retain direction and final decisions.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with planning and documentation of installations, directing the physical installation and coordinating hardware deployment across systems requires real-time decision-making, hardware troubleshooting, and coordination that current AI systems cannot reliably handle end-to-end without substantial human oversight and intervention.
Task automatabilityclaude-sonnet-52/5The task centers on directing/managing installations across systems, requiring coordination, judgment on sequencing, risk assessment, and stakeholder communication that current AI cannot fully replace, though scripted automation can handle sub-steps.
Adoption barriersclaude-haiku-4-5-202510014/5Organizations typically require a licensed or certified engineer to direct critical installations due to liability concerns, warranty implications, and the need to verify proper deployment before systems go live. Regulatory and organizational frameworks make fully autonomous direction legally and contractually risky.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but organizational risk tolerance, security compliance, and change-management protocols create friction against fully autonomous AI-directed installations in enterprise environments.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tool costs plus required human oversight and error correction are currently comparable to or exceed the cost of having an engineer direct installations, especially given the high cost of installation failures in production environments.
Cost vs. human wageclaude-sonnet-52/5While automation scripts reduce marginal costs for repeated deployments, the architectural direction and oversight role still requires paid engineering time, and AI tooling integration/maintenance costs are non-trivial relative to the task's supervisory nature.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably directs full installation workflows autonomously. AI tools can help with scripting and planning, but production installation direction still depends on human engineers for hardware connectivity issues, configuration validation, and real-time problem-solving.
Technical feasibility todayclaude-sonnet-52/5Deployed tools (Ansible, Terraform, MDM systems) automate portions of software/OS deployment, but 'directing' installation across heterogeneous environments still relies on human architects for planning and oversight; no product autonomously directs full installations reliably.

Establish functional or system standards to address operational requirements, quality requirements, and design constraints.

30

CI 28–32 · exposure 25 · augmentation 63 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Tech companies are experimenting with AI-assisted documentation and design validation, but actual adoption of AI for autonomous standard-setting is in the pilot phase. Most organizations still rely on human architects for final standard approval.
Sector adoption velocityclaude-sonnet-53/5Software/IT architecture roles are in a sector with moderate-to-fast AI tool adoption (copilots, design assistants), though full standards-setting work is largely still human-led with AI as auxiliary support.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by generating documentation templates, analyzing requirements for gaps, suggesting design patterns, and automating compliance checks. These augmentations improve speed and completeness, though the engineer remains responsible for judgment and approval.
Augmentation potentialclaude-sonnet-54/5AI tools can meaningfully assist by researching best practices, drafting standard templates, comparing design constraint tradeoffs, and generating documentation, substantially speeding up part of this task.
Task automatabilityclaude-haiku-4-5-202510012/5Establishing standards requires understanding complex operational requirements, quality constraints, and design tradeoffs that demand substantial human judgment and stakeholder alignment. While AI can draft candidate standards or suggest templates, the core task of arbitrating requirements and setting organizational standards remains inherently human-centered and cannot achieve 50% time savings end-to-end.
Task automatabilityclaude-sonnet-52/5This requires senior judgment about organizational context, tradeoffs, and stakeholder needs that current AI cannot autonomously determine; AI can draft standards documents but cannot independently establish authoritative standards without significant human direction.
Adoption barriersclaude-haiku-4-5-202510014/5Organizational standards often require formal sign-off from senior architects or governance bodies, and liability for poor standards can be significant. Regulatory frameworks in certain domains (safety-critical systems, finance, healthcare) may mandate human accountability in standard-setting.
Adoption barriersclaude-sonnet-53/5No licensing requirement typically, but organizational accountability, liability for system failures, and need for senior sign-off create meaningful friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for documentation and analysis are relatively cheap, but the task requires senior engineer involvement for validation, stakeholder alignment, and legal/compliance review, making the all-in cost comparable to or higher than unaugmented human effort.
Cost vs. human wageclaude-sonnet-52/5AI can reduce drafting time but the human architect still needs to define requirements, validate constraints, and take ownership, so overall cost savings are modest rather than order-of-magnitude.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed product reliably establishes functional or system standards autonomously. AI can assist with documentation and analysis, but standard-setting requires organizational authority, legal review, and domain expertise that deployed systems do not perform at production scale.
Technical feasibility todayclaude-sonnet-52/5Products exist that can generate draft technical standards or documentation given detailed specs, but no deployed system reliably establishes system standards independently in production architecture work.

Collaborate with engineers or software developers to select appropriate design solutions or ensure the compatibility of system components.

29

CI 25–32 · exposure 25 · augmentation 75 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Although software development has seen early AI adoption (code completion, documentation), architectural collaboration—especially cross-team design reviews and compatibility decisions—remains deeply human-centered and embedded in synchronous workflows. Displacement of this task is minimal; most deployments are pilots or assistive only.
Sector adoption velocityclaude-sonnet-53/5Software/IT engineering sectors show above-average AI tool adoption (copilots, code analysis), but architecture-level collaborative decision-making remains largely human-driven in practice.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist engineers by rapidly generating alternative design solutions, analyzing component compatibility matrices, surfacing precedents from prior projects, and drafting compatibility reports. These capabilities boost the engineer's productivity and breadth of exploration while the engineer retains decision authority.
Augmentation potentialclaude-sonnet-54/5AI coding assistants and system analysis tools can meaningfully speed up compatibility checking, documentation, and option comparison, improving engineer productivity while humans retain decision authority.
Task automatabilityclaude-haiku-4-5-202510012/5This task requires evaluating trade-offs, understanding domain context, and making judgments about compatibility in conversation with other professionals. While AI can assist in identifying candidate solutions or documenting compatibility checks, it cannot reliably lead the collaborative decision-making process that defines the task, nor replace the negotiation and consensus-building inherent to 'collaborating with engineers.'
Task automatabilityclaude-sonnet-52/5This is a collaborative, judgment-heavy task involving negotiation across stakeholders and architectural tradeoffs; AI can support parts (compatibility checks, documentation) but cannot autonomously conduct the collaboration or make final integration decisions.
Adoption barriersclaude-haiku-4-5-202510014/5Strong organizational and professional norms require that system design decisions and compatibility validation be reviewed and signed off by licensed or credentialed engineers. Liability for architectural decisions, regulatory compliance in critical domains, and contractual obligations to clients create hard barriers to full automation of this collaborative task.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but organizational accountability for system architecture decisions and liability for integration failures create meaningful friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5The per-task cost of AI-assisted design analysis (inference + tool integration) remains modest, but the task's complexity and need for human validation mean oversight costs are substantial. The all-in cost per decision remains comparable to or higher than that of a mid-level engineer spending focused time on the collaboration.
Cost vs. human wageclaude-sonnet-52/5AI tools can reduce some analysis time cheaply, but the core task requires senior engineer judgment and stakeholder communication, keeping human cost dominant and AI a supplement rather than substitute.
Technical feasibility todayclaude-haiku-4-5-202510012/5Current AI systems can generate design suggestions and compatibility analyses from documentation, but no deployed product reliably participates in the real-time, context-dependent collaboration and design review cycles typical in engineering teams. Tools exist to assist (e.g., code analysis, design templates), but they fall short of autonomous task execution.
Technical feasibility todayclaude-sonnet-52/5Products exist for code analysis, dependency checking, and design documentation assistance, but no deployed system reliably manages cross-team design collaboration or compatibility arbitration at production scale.

Define and analyze objectives, scope, issues, or organizational impact of information systems.

29

CI 25–32 · exposure 25 · augmentation 63 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI for autonomous strategic architecture and impact analysis remains limited; while larger tech and finance firms pilot AI-assisted requirements gathering and documentation, production displacement is minimal because these tasks demand human judgment and accountability that organizations are reluctant to delegate to automated systems.
Sector adoption velocityclaude-sonnet-53/5IT and professional services sectors are adopting AI for documentation and analysis support, but strategic systems architecture work still sees mostly pilot-stage augmentation rather than deep automation.
Augmentation potentialclaude-haiku-4-5-202510013/5AI provides useful assistance through automated documentation generation, requirements traceability, impact simulation, and preliminary stakeholder analysis, but the core task of defining objectives and assessing organizational consequences remains human-centered, with AI serving as a productivity aid rather than a transformative tool.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist by synthesizing requirements documents, generating draft scope statements, and analyzing impact scenarios, significantly speeding up parts of this task while humans retain final judgment.
Task automatabilityclaude-haiku-4-5-202510012/5Defining and analyzing strategic objectives, scope, and organizational impact requires deep contextual understanding of business goals, stakeholder needs, and system constraints—tasks where AI lacks the requisite domain expertise and cannot reliably produce end-to-end analysis without extensive human oversight. While AI can assist in data gathering and preliminary scoping, the synthesis and judgment required for high-stakes architectural decisions remains predominantly human work.
Task automatabilityclaude-sonnet-52/5This requires synthesizing organizational context, stakeholder priorities, and business judgment that current AI cannot reliably perform end-to-end without extensive human framing and validation.'
Adoption barriersclaude-haiku-4-5-202510014/5Strong organizational barriers exist: systems architecture decisions carry significant liability and financial consequences, requiring sign-off by licensed or credentialed engineers; regulatory frameworks in finance, healthcare, and critical infrastructure mandate human accountability for system design decisions, and organizational governance structures typically require human judgment and accountability for scope and impact assessments.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but organizational trust, accountability for strategic decisions, and need for stakeholder buy-in create moderate friction against full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools (LLMs, code analysis systems) have modest per-use costs but still require senior engineer oversight and validation, making the all-in cost per completed analysis substantial; the value of avoiding human error in this high-stakes domain often justifies the cost of human expertise rather than AI substitution.
Cost vs. human wageclaude-sonnet-52/5Human architects' judgment and stakeholder engagement are essential; AI assistance reduces some drafting time but does not replace the core analytical labor, so cost savings are modest.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed products reliably perform end-to-end definition and analysis of IS objectives and organizational impact in production environments; tools exist for requirements documentation and impact modeling, but they operate as assistants requiring expert validation rather than as autonomous decision-makers. AI systems lack the organizational context and decision authority to independently determine scope and impact assessment.
Technical feasibility todayclaude-sonnet-52/5AI tools can assist with drafting scope documents or summarizing requirements, but no deployed product independently defines system objectives or assesses organizational impact reliably.

Communicate with staff or clients to understand specific system requirements.

27

CI 21–32 · exposure 25 · augmentation 75 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI for requirements-gathering remains limited to narrow, well-structured scenarios (templated forms, simple questionnaires); most engineering organizations continue to rely on human architects and engineers for direct client interaction, with slow pilot activity.
Sector adoption velocityclaude-sonnet-53/5IT and professional services sectors are adopting AI meeting assistants and note-summarization tools moderately, but full automation of requirements discussions remains uncommon in production.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist engineers by auto-generating requirement summaries, flagging ambiguities, offering templates, and organizing client feedback—substantially raising human productivity in documentation and analysis while the engineer retains client relationship ownership.
Augmentation potentialclaude-sonnet-54/5AI can transcribe, summarize, and help structure requirements documents, draft follow-up questions, and organize stakeholder input, meaningfully boosting engineer productivity while humans still lead the conversation.
Task automatabilityclaude-haiku-4-5-202510012/5AI can draft summaries of requirements or suggest clarifying questions, but cannot reliably conduct natural two-way dialogue to disambiguate client needs, build trust, or navigate interpersonal nuance required for accurate requirement-gathering. Meaningful automation would require near-perfect conversational AI and domain context understanding—far below the 50% time-saving bar.
Task automatabilityclaude-sonnet-52/5Requirement gathering relies on nuanced human dialogue, reading unstated needs, and relationship building that current AI cannot fully replicate end-to-end, though AI can assist with note-taking and question generation.
Adoption barriersclaude-haiku-4-5-202510014/5Client relationships, trust-building, and accountability for system specifications typically require human contact and sign-off; organizational culture and client preference for direct engineer engagement create strong friction against substitution, even where technical capability might improve.
Adoption barriersclaude-sonnet-53/5No licensing requirement, but clients typically expect human engagement for trust, negotiation, and accountability in defining system scope, creating moderate organizational friction against full substitution.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of AI systems (inference, integration, validation oversight) combined with inevitable human re-work on missed or misunderstood requirements exceeds the loaded cost of a human engineer directly engaging with clients, making full automation uneconomical.
Cost vs. human wageclaude-sonnet-52/5Human engineers still must conduct most stakeholder conversations; AI tools reduce documentation time but don't replace the interactive cost of the core task, so savings are modest.
Technical feasibility todayclaude-haiku-4-5-202510012/5While chatbots can participate in scripted requirement conversations, no deployed product reliably captures complex, evolving system requirements from novel clients without significant human correction and refinement. Research prototypes exist but production systems consistently underperform on this nuanced communication task.
Technical feasibility todayclaude-sonnet-52/5Chatbots and AI meeting assistants can transcribe and summarize conversations, but no deployed product independently conducts full requirements-elicitation with clients reliably at scale.

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.