Statistical Assistants
43-9111.00Median wage $50,330/yr4,710 employed (US)Rank #12 of 923 scored · top 1% by substitution
Compile and compute data according to statistical formulas for use in statistical studies. May perform actuarial computations and compile charts and graphs for use by actuaries. Includes actuarial clerks.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution72
Exposure70
Augmentation78
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
14 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
71%
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%71
panel mean rating 3.8/5 → substitution pressure 71/100
Technical feasibility todayw 20%69
panel mean rating 3.7/5 → substitution pressure 69/100
Cost vs. human wagew 15%79
panel mean rating 4.1/5 → substitution pressure 79/100
Adoption barriersw 20%inverted — strong barriers lower the score77
panel mean rating 1.9/5 (barrier strength) → substitution pressure 77/100
Sector adoption velocityw 10%60
panel mean rating 3.4/5 → substitution pressure 60/100
Task breakdown (14 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.
Enter data into computers for use in analyses or reports.
97CI 97–97 · exposure 100 · augmentation 75 · importance 4.4/5 · click for rater detail
Enter data into computers for use in analyses or reports.
97| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Organizations across finance, healthcare, government, and professional services are actively deploying RPA and AI-based data entry automation; this is mainstream, not experimental, with documented displacement of entry-level roles. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Data entry automation via RPA and ETL tools is already widely deployed across finance, healthcare administration, and research support functions. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly assists humans by auto-completing forms, flagging inconsistencies, and pre-filling fields from documents, raising productivity of remaining manual work and allowing focus on validation and complex cases. |
| Augmentation potential | claude-sonnet-5 | 4/5 | Even where full automation is incomplete due to messy source formats, AI tools like OCR and autofill significantly speed up human data entry work. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Data entry into computers is a well-defined, rule-based task that current AI systems with OCR, form-filling, and database integration capabilities can perform end-to-end with >50% time savings at equal quality, especially with structured input formats. |
| Task automatability | claude-sonnet-5 | 5/5 | Data entry from structured or semi-structured sources is highly routinized and can largely be automated via OCR, APIs, form-parsing, and scripting tools that meet or exceed the 50% time-saving threshold. |
| Adoption barriers | claude-haiku-4-5-20251001 | 1/5 | No legal or regulatory barriers prevent automation of routine data entry; it is a clerical task with no licensing requirement, and organizational adoption is limited only by integration effort and change management, not hard constraints. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, liability, or human-contact requirements meaningfully block automating routine data entry. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | The all-in cost of AI-driven data entry (including inference, integration, and minimal oversight) is typically an order of magnitude cheaper than paying a human assistant for the same volume of data entry work. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated data entry pipelines cost a small fraction of a human's hourly wage per record processed, especially at volume. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature products exist today (RPA platforms, OCR+automation tools, AI-powered data extraction) that reliably perform data entry tasks in production across finance, healthcare, and government sectors at scale. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Mature production tools (RPA platforms, OCR/ETL pipelines, spreadsheet automation) reliably perform data entry and ingestion tasks at scale across industries today. |
Send out surveys.
96CI 91–100 · exposure 92 · augmentation 75 · importance 3.5/5 · click for rater detail
Send out surveys.
96| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 5/5 | Survey distribution automation is deeply embedded in information, finance, healthcare, and professional services sectors; automated survey platforms are the industry standard and have been for over a decade. |
| Sector adoption velocity | claude-sonnet-5 | 5/5 | Survey distribution automation is a long-established, deeply adopted practice across market research, HR, and academic/professional services sectors. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI augmentation is high for survey design review, recipient targeting, response tracking, and follow-up scheduling, though the human's role narrows significantly since the core distribution task is fully automatable. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI/automation tools substantially reduce the manual burden of survey distribution, letting assistants focus on design, targeting, and analysis, though some oversight remains useful. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can fully automate survey distribution via email, SMS, or web platforms with minimal setup, easily achieving 50%+ time savings and equal or better quality through systematic scheduling and tracking. The task lacks complexity beyond distribution logistics, which modern tools handle reliably. |
| Task automatability | claude-sonnet-5 | 5/5 | Sending surveys is a highly structured, repetitive administrative task easily handled by survey platforms and email automation tools with equal or better quality than manual dispatch. |
| Adoption barriers | claude-haiku-4-5-20251001 | 1/5 | No licensing, regulatory, or legal barriers prevent survey distribution automation; organizations widely adopt these tools without human sign-off requirements or liability concerns for the distribution task itself. |
| Adoption barriers | claude-sonnet-5 | 1/5 | No licensing, regulatory, or human-contact requirements restrict who or what can send surveys; organizations already automate this widely. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated survey distribution via established platforms costs pennies per survey at scale, representing a 10–100x cost advantage over manually sending surveys or managing distribution, even accounting for platform fees and minimal oversight. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated distribution tools cost a fraction of a cent per recipient compared to the labor cost of a human manually sending each survey. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Deployed survey platforms (Qualtrics, SurveyMonkey, Typeform) and marketing automation tools (HubSpot, Mailchimp) routinely handle survey distribution at scale in production environments across thousands of organizations daily. |
| Technical feasibility today | claude-sonnet-5 | 5/5 | Mature, widely deployed products (Qualtrics, SurveyMonkey, Google Forms, email marketing platforms) reliably automate survey distribution at scale in production today. |
Code data prior to computer entry, using lists of codes.
84CI 76–92 · exposure 83 · augmentation 75 · importance 4.1/5 · click for rater detail
Code data prior to computer entry, using lists of codes.
84| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Data-heavy sectors (finance, research, government) are experimenting with automated coding, but adoption remains piecemeal. Many organizations still rely on manual or legacy systems, reflecting moderate, uneven uptake rather than rapid, sector-wide deployment. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Data processing and statistical support functions are within professional/administrative services, a sector with fast digitization and widespread deployment of automated data classification tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments human coders by pre-suggesting or batch-coding candidates for review, reducing manual repetition and increasing output velocity. Humans remain in the QA loop, but their productivity is substantially elevated. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools substantially speed up human coders by pre-labeling or suggesting codes, letting humans focus on verification and exceptions, meaningfully raising throughput. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | This task is highly automatable. AI systems can recognize patterns, apply mapping rules from code lists to data, and perform end-to-end coding with minimal setup. OCR and structured data matching would enable >50% time savings at equal quality in most cases. |
| Task automatability | claude-sonnet-5 | 5/5 | Coding data against defined lookup lists is a well-structured, rule-based classification task that current AI (LLMs or even simpler ML classifiers) can perform end-to-end with substantial time savings at equal or better accuracy. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal barriers exist; coding is clerical work without licensing requirements. However, data governance, quality control oversight, and organizational processes that validate coded output create moderate friction to full automation. |
| Adoption barriers | claude-sonnet-5 | 1/5 | There is no licensing, liability, or regulatory requirement mandating a human perform simple data coding; it's a low-stakes clerical task with minimal adoption friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-based data coding via cloud APIs or open-source OCR/NLP solutions costs orders of magnitude less per record than human statistical assistants, especially at scale, after modest integration overhead. |
| Cost vs. human wage | claude-sonnet-5 | 5/5 | Automated coding via scripts or LLM API calls costs a small fraction of a cent per record versus a human assistant's hourly wage, making AI dramatically cheaper at scale. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products reliably perform this task through APIs and document processing services that extract, classify, and code structured data. While edge cases exist, mature solutions handle standard coding workflows in production environments. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Production systems (data pipelines using NLP classifiers, RPA tools, LLM-based categorization) are widely deployed for coding survey/administrative data against code lists, though edge cases still require human review. |
Compile reports, charts, or graphs that describe and interpret findings of analyses.
84CI 75–92 · exposure 83 · augmentation 100 · importance 4.4/5 · click for rater detail
Compile reports, charts, or graphs that describe and interpret findings of analyses.
84| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Information, finance, and professional services sectors are actively deploying BI dashboards and AI-driven report generation; adoption is rapid in digitized organizations. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Statistical/analytics work sits within professional services and data-heavy sectors that have shown fast uptake of AI-assisted reporting and BI tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI assistants substantially enhance productivity by drafting charts, generating narrative interpretations, and suggesting visualizations while analysts focus on deeper insight and validation. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially speeds up drafting narrative interpretations and generating visualizations while the assistant retains responsibility for accuracy and context, a strong augmentation case. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Current AI systems can generate reports, charts, and graphs end-to-end from structured data and analysis results, with off-the-shelf tools (e.g., large language models, visualization libraries) achieving >50% time savings at equal quality for routine compilations and interpretations. |
| Task automatability | claude-sonnet-5 | 4/5 | AI tools can generate charts, graphs, and narrative interpretations from structured data with minimal setup, meeting the time-saving threshold for most routine reporting tasks.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 1/5 | No licensing, regulatory, or liability barriers prevent automation; reports are typically reviewed by supervisory staff, and no legal requirement mandates human compilation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement typically governs producing internal reports/charts, though some regulated industries may require sign-off on interpretations for compliance purposes. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | AI-driven report generation (API inference + automation) costs a fraction of a statistical assistant's hourly wage, with minimal oversight needed for standard compilations. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Generating standard reports and visualizations via AI tools costs a small fraction of an analyst's hourly wage, though some oversight and data-cleaning cost remains. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature products (Python/R automation, BI tools, LLM-based report generators) reliably produce charts, graphs, and narrative reports in production, though reliability for nuanced interpretation of novel findings remains variable. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Deployed products (BI tools with AI copilots, code-generation assistants for data visualization, LLM-based report drafting) reliably produce charts and summary interpretations in production settings today. |
File data and related information, and maintain and update databases.
84CI 75–92 · exposure 87 · augmentation 63 · importance 4.2/5 · click for rater detail
File data and related information, and maintain and update databases.
84| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Banking, insurance, healthcare, and large enterprises have rapidly deployed RPA and database automation for these tasks over the past 5–10 years; adoption is deep and measured in production systems, though smaller or less digitized firms lag. |
| Sector adoption velocity | claude-sonnet-5 | 4/5 | Database automation and data pipeline tools are already deeply embedded in most information-processing and administrative workflows, reflecting fast, broad adoption patterns typical of office/clerical automation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI assists humans by automating routine filing and flagging data quality issues, allowing assistants to focus on validation and exception handling, though the task itself is primarily automatable rather than augmentation-centric. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI and automation tools substantially speed up data entry, validation, and database updates, letting human assistants focus on exceptions and quality control rather than rote transcription. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Filing data and maintaining databases are highly structured, rule-based tasks that current AI systems and robotic process automation (RPA) handle end-to-end with significant time savings. Data entry, validation, and database updates meet or exceed the 50% time-saving threshold with established tools like UiPath, Automation Anywhere, or custom AI pipelines. |
| Task automatability | claude-sonnet-5 | 4/5 | Filing, organizing, and updating structured data is highly routinized and well within current AI/automation capabilities, especially with scripting, ETL tools, and AI-assisted data entry.4This includes validation and formatting, though some edge cases and data quality judgment calls remain. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | While some organizations impose internal oversight or audit requirements for database changes and access controls exist, there are no legal licensing mandates requiring a human to perform or sign off on data filing itself. Integration into existing systems requires some technical configuration but encounters modest organizational friction. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal requirement mandates human execution; the main friction is organizational (data governance policies, legacy systems integration) rather than regulatory or liability-based. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated data filing and database maintenance cost orders of magnitude less than human labor when amortized across volume; a single RPA or cloud automation solution handles work that would require multiple human assistants, making the cost ratio heavily favorable to automation. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated ETL/database update systems and RPA tools cost a small fraction of a human assistant's wage per unit of throughput, though initial integration and oversight costs reduce the ratio somewhat from the maximum. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Mature products and deployed systems in organizations worldwide reliably perform data filing, entry validation, and database maintenance at scale. Cloud databases, RPA platforms, and AI-driven ETL tools demonstrate production-grade reliability for these exact tasks. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Database management tools, RPA bots, and AI-assisted data pipelines are widely deployed in production today for exactly this kind of maintenance and filing task, though full autonomy on unstructured or ambiguous data is less mature. |
Compute and analyze data, using statistical formulas and computers or calculators.
82CI 72–92 · exposure 87 · augmentation 100 · importance 4.5/5 · click for rater detail
Compute and analyze data, using statistical formulas and computers or calculators.
82| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Statistical automation is already deeply embedded in information, finance, and professional services. Organizations routinely use automated statistical software, dashboards, and data pipelines; adoption of AI-assisted statistical analysis is measurably accelerating in these digitized sectors. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Data analysis roles across research, business, and government are adopting statistical software and AI tools steadily, though full agentic automation of analysis pipelines remains uneven. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI-powered data analysis tools (ChatGPT with data, Copilot, domain-specific analytics platforms) meaningfully augment human statisticians by automating formula application, suggesting analyses, and generating visualizations while the human directs strategy and interpretation—transforming productivity while keeping humans central. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI and statistical software dramatically boost productivity for assistants performing computations, handling formula application, error-checking, and interpretation support while humans validate results. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | Statistical computation and basic analysis using formulas are directly automatable end-to-end with modern tools (Python, R, Excel macros, AI agents). These tasks involve deterministic mathematical operations that current AI systems execute reliably and faster than humans at equal quality, easily meeting the ≥50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 4/5 | Standard statistical computation and analysis with defined formulas is highly automatable using existing software, scripting, and AI-assisted code generation, saving significant time on routine work.stics. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few legal or licensing barriers exist for automating pure statistical computation. The main friction points are organizational adoption, quality oversight, and preference for human interpretation of results—but these are soft barriers, not hard regulatory or liability blockers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for this specific task, though organizational review of analysis quality and data handling introduces some friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Inference cost for statistical computation is minimal; integration into existing workflows is straightforward; oversight is light. AI automation is orders of magnitude cheaper than the loaded wage of a statistical assistant performing routine calculations and basic analysis. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated statistical computation via software/AI tools costs a small fraction of a human assistant's wage for equivalent routine calculations. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 5/5 | Deployed products (statistical software, spreadsheet automation, AI-powered data analysis tools like those in Power BI, Tableau, and Python libraries) reliably perform statistical computation and formula-based analysis in production at scale across organizations today. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Mature statistical software (R, Python, SPSS, Excel) plus AI coding assistants reliably perform standard computations in production today, though complex custom analyses still need human oversight. |
Check survey responses for errors, such as the use of pens instead of pencils, and set aside response forms that cannot be used.
81CI 72–90 · exposure 83 · augmentation 63 · importance 4.2/5 · click for rater detail
Check survey responses for errors, such as the use of pens instead of pencils, and set aside response forms that cannot be used.
81| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Government and survey organizations are gradually automating document intake, but adoption remains uneven. Many smaller survey firms and public agencies still use manual review; pilot deployments are common, but production integration across the sector is not yet widespread. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Survey/market research and data-processing firms have moderately adopted automated form-checking tools, though many smaller operations still rely on manual review. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist human reviewers by pre-screening forms and flagging borderline cases for verification, reducing review time and catching obvious errors automatically. However, the task itself is primarily mechanical, so augmentation gains are moderate compared to full end-to-end automation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI-based scanning tools can pre-flag likely errors, letting human assistants focus only on ambiguous or flagged cases, substantially boosting throughput. |
| Task automatability | claude-haiku-4-5-20251001 | 5/5 | This task involves straightforward pattern recognition and rule-based filtering: identifying physical format violations (pen vs. pencil) and marking unusable forms. Computer vision systems can reliably detect writing instrument type and image quality/legibility, and rule-based logic can route forms accordingly, achieving well over 50% time savings compared to manual review. |
| Task automatability | claude-sonnet-5 | 4/5 | Automated validation checks (e.g., OCR/OMR scanning systems, form-processing software) can flag pen use, incomplete responses, or unusable forms with high reliability, saving most manual review time.imonos.rasa Some edge cases still need human judgment for ambiguous forms. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | No legal mandate requires a human to perform this check; it is routine quality control. Minimal regulatory friction or liability asymmetry—errors simply trigger re-administration. Organizational inertia and staff retraining are the main frictions, not hard barriers. |
| Adoption barriers | claude-sonnet-5 | 1/5 | This is a routine clerical quality-check task with no licensing, liability, or regulatory requirement for human performance. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated document scanning and computer vision inference cost pennies per form, whereas a human statistical assistant checking forms manually costs $15–25+ per hour; even with oversight overhead, AI is at least 10–20× cheaper for high-volume survey intake. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated scanning/validation software processes large volumes of forms far cheaper per form than manual visual inspection by an assistant. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Document scanning and quality-control systems with OCR and image analysis are in production use across survey processing and government agencies. While some edge cases (ambiguous writing tools, marginal legibility) require human judgment, deployed systems reliably perform the core filtering task at scale with minimal oversight. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Scanning and optical mark recognition systems are widely deployed in survey processing pipelines and reliably flag formatting errors and unusable forms in production today. |
Check source data to verify completeness and accuracy.
76CI 72–79 · exposure 75 · augmentation 100 · importance 4.4/5 · click for rater detail
Check source data to verify completeness and accuracy.
76| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Data quality automation is widely adopted in information-heavy sectors (finance, tech, analytics, healthcare); major cloud platforms (AWS, Azure, GCP) embed data validation, and standalone tools see strong uptake. Deployment is both faster and deeper than many analytical tasks. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Data quality automation is common in finance and analytics-heavy sectors, but many organizations still rely on manual spot-checks, giving a mixed adoption picture. |
| Augmentation potential | claude-haiku-4-5-20251001 | 5/5 | AI augmentation is exceptionally strong here: automated flagging of anomalies, completeness summaries, and interactive dashboards allow humans to focus on exception handling and domain judgment rather than manual spot-checking, dramatically raising productivity. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI-driven anomaly detection and automated flagging dramatically speed up human review of source data, letting assistants focus only on flagged exceptions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Current AI systems can automate substantial portions of data validation work—detecting missing values, inconsistent formatting, outliers, and schema violations—with high accuracy. While domain-specific anomalies may require human judgment, the core verification logic is highly automatable and can save >50% of time at comparable quality. |
| Task automatability | claude-sonnet-5 | 4/5 | Data validation against defined rules (completeness checks, range checks, cross-field consistency, duplicate detection) is well within current AI and scripted tooling capabilities, meeting the time-saving threshold for most structured datasets. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or legal barriers prevent automation of data completeness and accuracy checking. Integration into data pipelines faces standard organizational friction and oversight requirements, but no licensing or human-signature mandate applies to this technical task. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement, but organizations may still want human sign-off on data quality for compliance-sensitive datasets, creating mild friction. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated data validation incurs minimal per-run cost (cloud compute for scanning) compared to the loaded wage of a statistical assistant performing manual checks for hours. AI-driven solutions are orders of magnitude cheaper at scale once infrastructure is amortized. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated validation scripts and AI tools run at a fraction of the cost of manual review per record, especially at scale, though initial rule-setup and oversight add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Data validation platforms and AI-driven data quality tools are deployed in production across finance, healthcare, and enterprise settings. Tools like Great Expectations, Soda, and cloud-native data pipelines reliably perform automated checks; accuracy is high though edge cases and complex business rule validation may still require manual review. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Data validation software, ETL pipeline checks, and AI-based anomaly detection tools are widely deployed in production for data quality assurance, though edge cases and domain-specific accuracy checks still need human review. |
Organize paperwork, such as survey forms or reports, for distribution or analysis.
76CI 67–84 · exposure 70 · augmentation 63 · importance 4.2/5 · click for rater detail
Organize paperwork, such as survey forms or reports, for distribution or analysis.
76| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Digital organizations, financial services, healthcare, and large government agencies have rapidly adopted document automation and RPA tools for paperwork organization over the past decade, with widespread production deployment. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Administrative/clerical support functions are adopting document automation tools steadily, but many organizations still rely on manual processes for paperwork handling, especially in survey/statistical agencies. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI-powered document classification and organization systems can assist humans in quickly routing and categorizing survey forms or reports, but the task itself is already largely automatable without human augmentation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can meaningfully speed up sorting, tagging, and pre-organizing documents for human review, improving throughput while a person still oversees quality and edge cases. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Organizing and distributing paperwork can be largely automated today using document management systems, OCR, and workflow automation tools that handle sorting, filing, and distribution with minimal human intervention, achieving well over 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 4/5 | Sorting, classifying, and preparing forms/reports for analysis is a structured, rules-based task that AI-driven document processing and OCR/classification tools can largely handle with modest setup. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few hard barriers exist; some organizations have legacy paper workflows or require human sign-off on sensitive distributions, but no legal licensing requirement prevents automation of organizing and distributing paperwork. |
| Adoption barriers | claude-sonnet-5 | 1/5 | This is a low-stakes clerical task with no licensing, liability, or human-contact requirements blocking automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 5/5 | Automated document organization and distribution costs a small fraction of the human labor required, especially at volume; inference and system integration are minimal compared to loaded wages for clerical staff performing this repetitive task. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Once digitized, automated sorting/tagging/routing via AI is far cheaper per unit than manual clerical labor, though initial digitization and integration add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Mature enterprise document management and workflow automation products (e.g., RPA, intelligent document processing platforms) are deployed in production across many organizations to organize, classify, and distribute documents reliably at scale. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Document management and intelligent document processing products exist and are used in production, but organizing heterogeneous paperwork (varied formats, handwriting, physical documents) still often requires human review to catch errors. |
Compile statistics from source materials, such as production or sales records, quality-control or test records, time sheets, or survey sheets.
72CI 72–72 · exposure 75 · augmentation 88 · importance 4.1/5 · click for rater detail
Compile statistics from source materials, such as production or sales records, quality-control or test records, time sheets, or survey sheets.
72| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Manufacturing, quality control, and finance sectors are adopting automation for record compilation at moderate pace through RPA and business process outsourcing, but adoption is not as rapid or comprehensive as in fully digital knowledge work domains. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Data compilation automation is common in finance and manufacturing quality control but adoption varies widely by sector size and digitization maturity, with many smaller operations still manual. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems can assist statistical assistants by automating data extraction and preprocessing while humans focus on validation, interpretation, and flagging anomalies—substantially raising human productivity on the compilation portion of the workflow. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI tools substantially speed up data compilation and error-checking, letting human assistants focus on validation and edge cases rather than manual tabulation. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | Extracting and aggregating numerical data from structured or semi-structured documents (records, sheets, surveys) is well within current AI capabilities. OCR, table extraction, and data consolidation can be largely automated, though complex unstructured source materials or error reconciliation may require some human oversight. |
| Task automatability | claude-sonnet-5 | 4/5 | Compiling statistics from structured or semi-structured source records is a data extraction and aggregation task that current AI/automation tools handle well, especially with templated inputs, though messy/unstructured sources still need setup or human validation."},"feasibility":{"rating":4,"rationale":"OCR, RPA, and LLM-based document parsing tools are deployed in production for compiling data from records like time sheets and survey forms, though edge cases and format variability still cause errors requiring review. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Few regulatory or authorization barriers exist for automated statistical compilation. Data governance and validation oversight may be required, but these are organizational rather than legal constraints on substitution. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing or legal requirement mandates a human compile these statistics; main friction is organizational trust in data accuracy and integration effort with legacy systems. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven data extraction and compilation costs are substantially lower than manual data entry and statistical compilation. A single AI process can handle volumes that would require multiple human assistants, yielding a clear cost advantage. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Automated data compilation via software is dramatically cheaper per record than manual tabulation once set up, though initial integration and occasional human QA add some cost. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Multiple deployed products (RPA tools, document intelligence APIs, spreadsheet automation agents) reliably extract and compile statistical data from production/sales records and forms in production settings. Success rates are high on standardized formats, though edge cases may still require human review. |
| Technical feasibility today | claude-sonnet-5 | 4/5 | Established RPA and document-extraction products (e.g., intelligent document processing tools) are widely used in production to compile statistics from structured business records, though accuracy on non-standard formats remains imperfect. |
Select statistical tests for analyzing data.
59CI 59–59 · exposure 50 · augmentation 88 · importance 4.2/5 · click for rater detail
Select statistical tests for analyzing data.
59| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Statistical software vendors and analytics platforms are piloting AI-assisted test selection; adoption is growing in academic and corporate research contexts. However, widespread production deployment remains modest, with many organizations still relying on manual selection or statistician consultation. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Data analysis and research support functions are seeing moderate AI tool adoption (e.g., in academia, research support, business analytics) but many organizations still rely on trained staff for methodology choices. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI significantly augments human statisticians by instantly proposing candidate tests, explaining assumptions, and flagging common pitfalls, allowing the human to validate and refine choices faster. This keeps humans in the loop while dramatically improving productivity on test selection. |
| Augmentation potential | claude-sonnet-5 | 5/5 | AI substantially speeds up brainstorming and narrowing statistical test options, letting human assistants validate and finalize choices much faster than working alone. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can suggest appropriate statistical tests given data characteristics and research questions, reducing setup time significantly. However, determining the correct test often requires domain expertise, assumption checking, and validation that typically still need human oversight to ensure statistical validity. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can suggest appropriate statistical tests given a described dataset and research question, often correctly for standard cases, but complex or ambiguous designs still require human judgment about assumptions and context. |
| Adoption barriers | claude-haiku-4-5-20251001 | 2/5 | Test selection requires human judgment and is typically not regulated or licensed. Organizational friction exists (preference for human statisticians on critical analyses), but nothing legally prevents AI recommendation systems or makes human sign-off mandatory. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No licensing requirement for choosing a statistical test, though downstream conclusions may face institutional review or publication scrutiny that indirectly requires human sign-off. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI inference cost for test selection is minimal compared to the loaded wage of a statistical assistant. Once integrated into workflows, the per-task cost is orders of magnitude lower than human labor, though integration overhead exists. |
| Cost vs. human wage | claude-sonnet-5 | 4/5 | Querying an LLM for test selection costs a fraction of a cent versus a human assistant's time, though some oversight and iteration is needed to validate the recommendation. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Multiple products (e.g., statistical software with AI suggestions, ChatGPT, specialized analytics platforms) can recommend tests based on data type and research goals. Performance is reliable for common scenarios but less robust for complex, multi-factorial designs or unusual data distributions. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Tools like ChatGPT, Copilot for Excel/R, and specialized stats assistants routinely recommend tests, but reliability drops for nuanced experimental designs, and no widely deployed production system autonomously handles this without human review. |
Participate in the publication of data or information.
47CI 39–55 · exposure 50 · augmentation 75 · importance 4.2/5 · click for rater detail
Participate in the publication of data or information.
47| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Statistical and research organizations historically adopt publishing automation slowly; while data pipelines are increasingly automated in tech-forward firms, publication tasks in government, academia, and regulated sectors remain heavily manual and human-controlled, reflecting laggard sectoral adoption patterns. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Statistical and data-reporting functions are increasingly using automation and AI-assisted drafting tools, though full-scale production replacement of publication workflows is still uneven across organizations. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI can substantially assist by auto-formatting data, generating publication-ready tables and figures, checking for consistency, and drafting metadata—tasks that consume significant time; a statistical assistant using such tools can publish more work and with fewer errors while maintaining oversight, representing meaningful productivity gain. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids drafting, summarization, and formatting of statistical outputs for publication, greatly increasing throughput even though human oversight remains essential for accuracy and compliance. |
| Task automatability | claude-haiku-4-5-20251001 | 3/5 | AI can automate portions of publication workflows—formatting data, generating summaries, and preparing outputs for distribution—but human judgment on data accuracy, regulatory compliance, and editorial decisions typically remains necessary, limiting time savings to roughly 30–50% depending on context. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft summaries, format tables, generate charts, and prepare reports for publication, but final review, sign-off, and coordination with stakeholders still require human involvement.redutime savings are partial rather than end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Publication often carries regulatory, compliance, and audit requirements (especially in healthcare, finance, government); liability for incorrect published data is typically assigned to human sign-off; organizational norms and legal frameworks often mandate human responsibility for data release, creating strong legal and reputational barriers to full automation. |
| Adoption barriers | claude-sonnet-5 | 2/5 | There are no licensing requirements for this task, though organizational data-quality standards and institutional review processes create some friction before publication. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for publication automation carry integration and oversight costs, and the need for human validation of published data means labor is not substantially displaced; total cost per publication cycle is often comparable to or higher than a statistical assistant's loaded wage, especially when liability for errors is factored in. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI tools reduce drafting and formatting time significantly, but the need for human review, quality checks, and domain-specific validation keeps blended costs only moderately lower than fully manual processes. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Deployed tools exist for data formatting, automated report generation, and document publishing (e.g., RPA, document-generation platforms), but they require significant configuration and human oversight; error rates in data integrity checks and content validation remain material enough that production use is typically narrow or requires substantial human review. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Products like automated reporting tools, BI dashboards, and LLM-based drafting assistants are deployed in production for data publication workflows, but they still require human editing and validation, limiting full reliability. |
Interview people and keep track of their responses.
39CI 34–45 · exposure 25 · augmentation 63 · importance 4.3/5 · click for rater detail
Interview people and keep track of their responses.
39| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 3/5 | Adoption is growing in market research and simple survey contexts, with major platforms experimenting with AI interviewers, but implementation remains scattered across the sector. Established survey firms are testing AI but have not yet displaced human interviewers at scale. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Survey and research organizations have been slow to adopt AI-led interviewing at scale, though chatbot-based intake tools are piloted in some market research and government survey contexts. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist human interviewers by auto-transcribing, auto-coding responses, flagging inconsistencies, and suggesting follow-up questions in real time, raising interviewer throughput and accuracy. However, the human still conducts the interview, limiting the productivity multiplier. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist with transcription, real-time note-taking, response coding, and follow-up prompt generation, significantly boosting interviewer efficiency while a human remains in control. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Current AI can conduct structured interviews and log responses in controlled settings, but struggles with nuanced follow-up, rapport, non-verbal cues, and adversarial or evasive respondents—the hallmarks of real fieldwork interviews. Significant human oversight remains necessary. |
| Task automatability | claude-sonnet-5 | 2/5 | Conducting interviews and probing for nuanced or unstructured responses still requires human judgment, rapport-building, and adaptive follow-up that AI struggles to fully replace, though structured survey portions could be automated.' |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | IRB approval, informed consent requirements, and data privacy regulations (GDPR, CCPA) create friction but do not legally forbid AI interviewers. Some respondents and organizations prefer human contact, and liability concerns around data handling introduce friction but not absolute barriers. |
| Adoption barriers | claude-sonnet-5 | 2/5 | No strict licensing requirement for most statistical interviewing, but survey validity, respondent trust, and data quality concerns create moderate organizational friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-based interview systems (voice bots, chatbots, or agent frameworks) have negligible per-call marginal cost compared to human interviewers who require wages, benefits, training, and commuting. Even accounting for integration and error correction, the cost advantage is substantial. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | For simple structured data collection AI-driven tools can be cheaper, but for full interviewing including rapport and adaptive questioning, human labor cost is comparable once quality and error correction are factored in. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While chatbots can conduct simple surveys or pre-scripted interviews at scale, production systems rarely handle the flexibility, context-switching, and relationship-building required in actual statistical interview work. Narrow-use products exist but lack reliability for diverse respondent populations. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Voice bots and chat-based survey tools exist for simple structured interviews, but reliable handling of open-ended, sensitive, or complex interview scenarios in production is limited and error-prone. |
Discuss data presentation requirements with clients.
24CI 16–32 · exposure 17 · augmentation 63 · importance 3.8/5 · click for rater detail
Discuss data presentation requirements with clients.
24| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | While professional services firms experiment with AI-assisted note-taking and drafting, client-facing discussions remain predominantly human-led. Adoption of AI for primary client engagement on data requirements is still in early pilots rather than scaled production deployment. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Professional/statistical services are adopting AI tools moderately for drafting and analysis support, but client-facing requirements discussions remain largely human-led with pilots only emerging. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by drafting meeting agendas, summarizing client feedback, suggesting visualization options, and preparing background materials—useful productivity aids. However, the core interpersonal negotiation and real-time adaptation still depend heavily on human judgment and presence. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can help prepare discussion materials, summarize prior conversations, suggest visualization options, and draft follow-up questions, meaningfully aiding the human during and around the discussion. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires nuanced two-way dialogue with clients about their specific needs, preferences, and constraints—understanding context, asking clarifying questions, and adapting explanations in real time. Current AI systems lack the sustained conversational and relational understanding needed to conduct genuine client discussions that would meet the 50% time-saving bar. |
| Task automatability | claude-sonnet-5 | 2/5 | This is a live client-facing consultation requiring real-time clarification of ambiguous needs, negotiation, and trust-building, which current AI cannot fully replace end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Client-facing advisory work carries significant reputational and relationship risk; organizations typically require human accountability and trust-building in client discussions. Liability concerns and client preference for human judgment on presentation strategy create substantial organizational and contractual friction against full substitution. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No licensing requirement, but client preference for human interaction, trust, and organizational norms around consultative work create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference costs for extended client conversations are modest, but meaningful oversight is required to ensure client satisfaction and correctness. When combined with necessary human validation and intervention, the all-in cost remains comparable to or potentially higher than hiring a statistical assistant directly. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Human discussion involves relationship management and judgment that AI substitutes poorly for, so despite low inference cost, the need for human oversight and correction keeps overall cost comparable to a human. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | While LLMs can generate text about data visualization principles and draft responses to client queries, no deployed system reliably conducts authentic client meetings or negotiations. Existing chatbots can simulate discussion but do not demonstrate the contextual judgment and relationship-building required in production client-facing work. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Chatbots and AI assistants can gather requirements via forms or scripted dialogue, but no deployed product reliably conducts nuanced client discovery conversations for statistical presentation needs at scale. |
Related occupations — Office & Administrative Support
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.