2026 05 01 srinivasan ai automation augmentation role replacement

Prof Suraj Srinivasan's automation and augmentation scores: which job roles will Artificial Intelligence replace entirely?

Research Question

What does Prof Suraj Srinivasan's research framework for measuring Automation and Augmentation (A&A) scores across job roles and industries reveal about which roles face full Artificial Intelligence (AI) replacement, specifically the finding that roles where automation score exceeds augmentation score are those AI will replace in totality, and what are the strategic implications for workforce and organisational planning?

Scope

In scope:

Prof Suraj Srinivasan's study, its methodology, and the specific infographic mapping industries and job roles along automation versus augmentation dimensions
The threshold criterion: roles where automation score > augmentation score are candidates for full AI replacement; roles where augmentation score > automation score are candidates for AI-assisted amplification of human capability
Industry and role breakdown from the infographic: which sectors and roles fall into each quadrant
The definitions Srinivasan uses for "automation" (AI performs the task end-to-end, reducing or eliminating the human role) and "augmentation" (AI assists a human who retains ownership of the task and judgment)
Strategic implications: what the automation-over-augmentation finding means for hiring, reskilling, and organisational design decisions

Out of scope:

Broad survey of all AI-and-jobs literature, this item is specifically bounded to Srinivasan's framework and infographic
Speculative long-run projections beyond what the study provides
Technical AI capability assessments not grounded in the study

Constraints:

Primary source must be the study itself (working paper, journal article, or Harvard Business Review (HBR) article by Srinivasan); do not synthesise from secondary summaries alone
Flag where findings rely on the infographic interpretation rather than formal written claims in the paper
The study must be dated and cited with a URL

Context

The accessible primary source for this item is Chen, Srinivasan, and Zakerinia's Working Paper 25-039, Displacement or Complementarity? The Labor Market Impact of Generative AI, supplemented by Harvard Business School Working Knowledge's February 2026 summary of the same research. [fact; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]

The paper's core contribution is not a universal claim that AI will replace whole occupations once one score exceeds another, but a task-composition framework that separates occupations dominated by automatable tasks from occupations that mix AI-exposed and non-exposed tasks in ways that enable human-AI collaboration. [inference; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]

That distinction matters for workforce planning because the evidence points to shrinking demand and skill simplification in structured clerical work, while judgment-heavy and mixed-task occupations see expanding demand for AI-related skills and broader role complexity. [fact; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]

Approach

Locate the primary source - Find the specific Srinivasan study (working paper, Harvard Business Review (HBR) article, or journal publication) that contains the automation/augmentation infographic. Verify authorship, date, and publication venue. Retrieve the full text.
Extract the methodology - How are automation and augmentation scores computed? What data sources are used (task taxonomies, Occupational Information Network (O*NET) occupational data, survey data, or other)? What is the unit of analysis, tasks, roles, or industries?
Map the infographic findings - Which industries and roles appear in the automation-dominant quadrant? Which appear in the augmentation-dominant quadrant? Reproduce the key data points in a structured evidence map.
Interrogate the threshold criterion - What justifies using automation > augmentation as the replacement threshold? Is this a formal model result or an interpretive framing? Are there caveats or uncertainty ranges?
Assess cross-study validation - Does the Srinivasan framework align with or diverge from comparable frameworks (e.g., McKinsey Global Institute occupation automation potential, Accenture "New Skills Now", World Economic Forum (WEF) Future of Jobs reports)? Note convergent and divergent findings.
Derive strategic implications - What should organisations do with this information? What is the recommended response for roles that score automation > augmentation? What does the augmentation zone recommend about human-AI teaming?

Sources

Research Skill Output

(Full output from running the research skill, retained verbatim. Sections 0-5 are the investigation; Section 6 seeds the Findings section below.)

§0 Initialise

[fact] Research question restated: Does Srinivasan's automation-versus-augmentation framework show which roles generative AI will replace entirely, and what should organizations do with that result? Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] Scope confirmed: The item is bounded to Srinivasan's framework, the accessible working paper, the Working Knowledge summary, the occupations and sector signals those sources expose, and cross-checks against official World Economic Forum (WEF) and McKinsey sources. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america
[fact] Primary evidence base: The strongest direct evidence comes from Working Paper 25-039, the Americas Conference on Information Systems proceedings abstract, and the February 2026 Harvard Business School Working Knowledge article summarizing an updated version of the paper. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://aisel.aisnet.org/amcis2025/sig_cnow/sig_cnow/18/; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[fact] Output format: The result will be a knowledge item with labelled evidence, explicit uncertainty, a structured synthesis, and a Findings section that does not introduce claims absent from Section 6. Source: https://github.com/davidamitchell/Research/blob/main/research-prompt.md

§1 Question Decomposition

Q: What does Srinivasan's framework actually show about replacement versus augmentation?
├── Q1: What are the primary sources, and what do they explicitly claim?
│   ├── Q1a: What does the December 2024 working paper say?
│   ├── Q1b: What does the February 2026 Working Knowledge summary add?
│   └── Q1c: Does the Americas Conference on Information Systems proceedings abstract corroborate the same framing?
├── Q2: How are automation and augmentation measured?
│   ├── Q2a: How is the automation index constructed from task exposure?
│   ├── Q2b: How is the augmentation index constructed from task mix?
│   └── Q2c: Are the two scores symmetric enough to justify a simple greater-than threshold?
├── Q3: Which occupations and sectors appear most exposed on each side?
│   ├── Q3a: Which roles top the automation ranking?
│   ├── Q3b: Which roles top the augmentation ranking?
│   └── Q3c: What sector-level evidence is explicitly stated?
├── Q4: Does the paper itself say that automation score > augmentation score means total replacement?
│   ├── Q4a: Is that threshold stated anywhere in the paper or summary?
│   └── Q4b: What is the strongest defensible interpretation instead?
├── Q5: How does this framework compare with broader labor-market frameworks?
│   ├── Q5a: Where does it align with the WEF Future of Jobs framing?
│   └── Q5b: Where does it align with McKinsey's augmentation-versus-displacement framing?
└── Q6: What are the strategic implications for workforce and organizational planning?
    ├── Q6a: What should firms do for automation-prone roles?
    └── Q6b: What should firms do for augmentation-prone roles?

§2 Investigation

A. Primary source identification

[fact] Working Paper 25-039, Displacement or Complementarity? The Labor Market Impact of Generative AI, is authored by Wilbur Xinyuan Chen, Suraj Srinivasan, and Saleh Zakerinia and is dated December 2024 in the accessible Harvard Business School PDF. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] The Americas Conference on Information Systems 2025 proceedings abstract states the same paper studies whether generative AI displaces workers or augments jobs by analyzing labor demand and skill requirements across occupations and finds a heterogeneous effect rather than a uniform replacement story. Source: https://aisel.aisnet.org/amcis2025/sig_cnow/sig_cnow/18/
[fact] Harvard Business School Working Knowledge's February 2026 article identifies the same study as the basis for an occupation-level visualization and reports updated labor-market effects through March 2025. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs

B. Methodology and score construction

[fact] The paper uses Occupational Information Network (O*NET) v25.1 task descriptions, the United States Department of Labor's occupational database, and Lightcast U.S. job-posting data, with 923 occupations and 19,265 tasks in the accessible working-paper version. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.onetonline.org/
[fact] Tasks are classified into four exposure buckets for generative-AI language capability: E0 no exposure, E1 direct exposure, E2 exposure through language-model-powered applications, and E3 exposure given image capabilities. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://research.ibm.com/blog/what-are-large-language-models
[fact] The automation index weights E1 tasks at 1, E2 tasks at 0.5, and E0 and E3 tasks at 0, with each task weighted by its O*NET importance score. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] The augmentation score is defined as 1 - (share of gen AI exposed tasks^2 + share of non-gen AI exposed tasks^2), again weighted by task importance, so it rises when an occupation mixes AI-exposed and non-exposed work rather than when it is fully exposed. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[inference] Because the augmentation score is a task-mix concentration measure and the automation index is a weighted exposure measure, the two scores are not mirror images on a common scale and are not designed for a single decisive crossover rule. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[inference] The augmentation formula peaks at 0.5 when exposed and non-exposed task shares are balanced, which means a raw comparison of automation and augmentation values is structurally asymmetric even before any empirical interpretation is added. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf

C. Occupation and sector signals

[fact] The top automation occupations in Table 3 are Correspondence Clerks, Interpreters and Translators, Court Municipal and License Clerks, Medical Transcriptionists, Telemarketers, Word Processors and Typists, Climate Change Policy Analysts, Order Clerks, and Payroll and Timekeeping Clerks. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] The top augmentation occupations in Table 3 are Clinical Neuropsychologists, Medical Dosimetrists, Fish and Game Wardens, Agricultural Engineers, Cartographers and Photogrammetrists, Traffic Technicians, Microbiologists, Arbitrators Mediators and Conciliators, and several supervisory or specialist roles. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] The Working Knowledge article adds concrete augmentation examples that are easier to interpret for managers, naming microbiologists, financial analysts, and clinical neuropsychologists as roles where AI can assist but human judgment remains central. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[fact] The same Working Knowledge article states that the largest reductions in job postings were in the finance and technology sectors. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[inference] The accessible paper gives stronger occupation-level evidence than sector-level mapping, so any full industry-by-quadrant reconstruction would go beyond what the paper tables directly expose. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs

D. Labor-demand and skills effects

[fact] In the accessible December 2024 working-paper version, occupations in the top quartile of automation exposure show a 17% decline in job postings per firm per quarter after ChatGPT's introduction, while occupations in the top quartile of augmentation exposure show a 22% increase. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] In the same accessible paper version, automation-prone occupations show declines in all required skills, AI-exposed skills, and new skills, while augmentation-prone occupations show increases in all three measures. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] The February 2026 Working Knowledge article reports updated magnitudes: a 13% decline in postings for structured repetitive occupations and 20% growth in more analytical, technical, or creative occupations, plus 7% fewer skills in automation-prone postings and more AI-related skills in augmentation-prone postings. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[fact] The authors' own managerial recommendation is to invest in reskilling for automation-prone occupations and continuous upskilling in generative AI for augmentation-prone occupations. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs

E. Threshold test: does automation > augmentation imply full replacement?

[fact] The accessible working paper does not state that any occupation with an automation score above its augmentation score will be replaced "in totality" by AI. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] The paper's empirical design treats occupations in the top quartile of automation exposure and the top quartile of augmentation exposure as separate treatment groups rather than defining one formal crossover threshold between the two indices. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[inference] The strongest supportable interpretation is that occupations dominated by structured, repetitive, cognitively codifiable tasks face contraction and deskilling pressure, while occupations with a balanced mix of automatable and non-automatable tasks are more likely to be redesigned around human-AI collaboration. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[inference] The phrase "replace entirely" overstates what the study measures because the dependent variables are job postings and skill requirements, not complete occupational disappearance or direct headcount elimination. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf

F. Cross-study validation

[fact] The World Economic Forum's Future of Jobs Report 2023 projects 23% labor-market churn over five years, with 69 million jobs created and 83 million eliminated, and identifies clerical or secretarial roles such as bank tellers, cashiers, ticket clerks, and data-entry clerks as among the fastest-declining roles. Source: https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf
[fact] The same WEF report says analytical thinking and creative thinking remain the most important worker skills, that 44% of workers' skills will be disrupted within five years, and that 61 out of 100 representative employees will require some form of training before 2027. Source: https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf
[fact] McKinsey's official 2023 future-of-work framing says generative AI is more likely to enhance how Science, Technology, Engineering, and Mathematics (STEM), creative, business, and legal professionals work than to eliminate large numbers of those jobs outright, while office support, customer service, and food service face the sharpest contraction risk. Source: https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier
[inference] Srinivasan's framework aligns with WEF and McKinsey on the direction of risk, routine clerical work contracts first and judgment-rich mixed-task work is more often augmented, but it adds a more explicit task-mix mechanism and uses job-posting evidence rather than executive surveys or macro projections. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america

§3 Reasoning

[fact] The paper directly supports a heterogeneous labor-market effect, not a single replacement outcome, because it shows demand and skill declines for automation-prone quartiles and demand and skill growth for augmentation-prone quartiles. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[inference] The numeric relation between automation and augmentation scores is weaker evidence than the occupational quartile results because the indices measure different constructs and were not presented as a formal decision rule by the authors. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[inference] A defensible reading is therefore "roles dominated by exposed repetitive tasks face contraction and possible role redesign," not "every role with automation above augmentation will disappear in full." Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs

§4 Consistency Check

[fact] The accessible working-paper version and the later Working Knowledge article disagree on both the sample end date and the headline effect sizes, with the paper using Lightcast data through June 2024 and reporting 17% down and 22% up, while the article says the dataset runs through March 2025 and reports 13% down and 20% up. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[inference] This looks like an updated sample window rather than a conceptual reversal, because the direction of results and the managerial recommendations remain consistent across both sources. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[fact] No accessible source located in this session states the proposed "automation score > augmentation score means total replacement" threshold in the authors' own words. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs

§5 Depth and Breadth Expansion

[inference] Economic lens: The study implies that the earliest labor-market effect of generative AI is fewer entry points into routine cognitive occupations, not instant mass unemployment, because job postings and required skills fall before the paper claims complete occupational elimination. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[inference] Behavioral lens: Augmentation-prone roles persist because firms still value human judgment, interpersonal interaction, and responsibility for consequential decisions even when AI can compress the routine parts of the workflow. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[inference] Organizational lens: Workforce planning should focus less on declaring whole job families "gone" and more on decomposing each role into automatable tasks, judgment tasks, and new AI-coordination tasks, then redesigning training and accountability around that mix. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf
[inference] Strategic lens: Firms that treat generative AI purely as a cost-cutting automation lever risk underinvesting in the occupations where AI raises skill complexity and productivity, which is the part of the labor market where the Srinivasan evidence shows demand growth. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america

§6 Synthesis

Executive summary:

[inference] Srinivasan's accessible research does not establish a formal rule that occupations with automation scores above augmentation scores will be replaced entirely by AI; it shows that high-automation occupations lose postings and skill breadth while high-augmentation occupations gain demand and AI-related skill requirements. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[fact] The framework combines an exposure-based automation index with a task-mix augmentation index, so the two measures are complementary but not a single binary decision boundary. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[fact] The occupations most exposed to automation are clerical, transcription, translation, and other structured cognitive roles, while the occupations most exposed to augmentation are mixed-task specialist roles such as clinical neuropsychologists, medical dosimetrists, agricultural engineers, and microbiologists. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[inference] Strategically, firms should plan for role redesign, reskilling, and selective headcount contraction in routine roles, while expanding AI literacy and human-AI collaboration in occupations where judgment and non-automatable tasks remain central. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf

Key findings:

[fact] The accessible Srinivasan research measures automation and augmentation as different constructs, so it does not present a formal crossover rule in which a higher automation score than augmentation score automatically means total occupational replacement. (medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
[fact] The working paper finds that occupations in the top quartile of automation exposure experienced a 17% decline in job postings per firm per quarter after ChatGPT, while occupations in the top quartile of augmentation exposure experienced a 22% increase. (medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
[fact] The later Harvard Business School Working Knowledge summary reports updated effects, including a 13% decline for structured repetitive occupations and 20% growth for more analytical, technical, or creative occupations. (medium confidence; source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs)
[fact] The most automation-exposed occupations are concentrated in clerical and codifiable language work, including correspondence clerks, interpreters and translators, court clerks, medical transcriptionists, telemarketers, typists, and payroll clerks. (medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
[fact] The most augmentation-exposed occupations are mixed-task specialist roles such as clinical neuropsychologists, medical dosimetrists, agricultural engineers, cartographers, microbiologists, and mediators, where AI can compress sub-tasks but not displace the need for human judgment or responsibility. (medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs)
[inference] The study supports a role-redesign thesis more than an occupation-extinction thesis, because its outcome variables are posting volume and skill composition rather than observed full elimination of whole job families. (medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
[inference] Srinivasan's results align with broader labor-market frameworks from the WEF and McKinsey, both of which also place the highest near-term risk on clerical or routine support work and place the strongest augmentation effects in higher-judgment knowledge work. (medium confidence; source: https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier)

Evidence map:

claim	source	confidence	notes
[fact] Automation and augmentation are defined through separate indices, not a published crossover rule.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	score construction
[fact] Top automation quartile postings fall 17% while top augmentation quartile postings rise 22% in the accessible paper version.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	Table 7
[fact] Updated Harvard Business School summary reports 13% decline and 20% growth.	https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs	medium	later sample window
[fact] Top automation occupations cluster in clerical and codifiable language work.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	Table 3
[fact] Top augmentation occupations cluster in mixed-task specialist roles.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs	medium	Table 3 plus manager-friendly examples
[inference] The evidence supports role redesign and contraction pressure more strongly than total occupational disappearance.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	dependent variables are postings and skills
[inference] WEF and McKinsey point in the same direction on routine clerical decline and judgment-rich augmentation.	https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier	medium	cross-study triangulation

Assumptions:

[assumption] The February 2026 Harvard Business School Working Knowledge article reflects a later revision of the same research program rather than a different model specification, because it cites the same authors, paper title, and qualitative conclusions while updating the sample window and headline percentages. Justification: the article explicitly says the paper was first released in December 2024 and later updated. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[assumption] The original infographic likely visualizes a larger set of occupations than the paper's published top-10 table, so the item relies on the table and article examples rather than claiming a complete reconstruction of every plotted point. Justification: the accessible sources expose ranked examples and sector notes but not a downloadable full coordinate list. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs

Analysis:

[inference] The paper traces how generative AI changes labor demand across occupations by linking task composition and skill mix to posting changes. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf
[inference] That matters strategically because a firm can automate large parts of a role without making the role disappear, especially where non-automatable decision rights, interpersonal accountability, or physical-world execution remain attached to the job. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[inference] The strongest workforce response is portfolio management across tasks and skills: reduce exposure in clerical and transcription-heavy work, redesign specialist mixed-task roles around AI assistance, and expand reskilling capacity before job-entry pathways narrow further. Source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf

Risks, gaps, uncertainties:

[fact] The accessible paper and the later Harvard Business School summary expose different sample end dates and different effect sizes, so the exact percentages should be treated as version-specific rather than timeless constants. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[fact] The accessible sources do not provide a full industry-by-role coordinate dump for the infographic, so the occupation mapping here is strongest for the published top-ranked roles and weaker for exhaustive quadrant reconstruction. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[inference] The McKinsey comparison is useful for directional triangulation but less precise than the Harvard Business School and WEF evidence in this item because this synthesis relies on McKinsey's official summary framing rather than line-by-line extraction of the underlying report text. Source: https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier

Open questions:

How far do the updated 2025 and 2026 versions of the Srinivasan research move the occupation rankings once more post-ChatGPT hiring data is included?
Can the full occupation-by-score dataset behind the Harvard Business School visualization be recovered from a public appendix or data release?
Which entry-level pathways are most vulnerable when postings fall in automation-prone occupations before unemployment visibly rises?

§7 Recursive Review

[fact] The central claims in this item are anchored in the accessible Harvard Business School working paper's score definitions, result tables, and occupation rankings, with the later Working Knowledge article used to capture the updated sample window and managerial recommendations. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs
[fact] Cross-study validation is anchored in official WEF and McKinsey URLs and is used only for directional comparison, not to override the paper's own definitions or findings. Source: https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier
[inference] The overall confidence remains medium because the stronger "replace entirely" threshold is not stated by the authors, the accessible sources do not fully expose the infographic dataset, and the updated Harvard Business School summary differs modestly from the accessible December 2024 paper version. Source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs

Findings

Executive Summary

Srinivasan's accessible research does not show that occupations with automation scores above augmentation scores will be replaced entirely by AI; it shows that occupations in the top automation quartile lose postings and skill breadth, while occupations in the top augmentation quartile gain demand and AI-related skill requirements. [inference; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]

The framework combines an exposure-based automation index with a task-mix augmentation index, so the two numbers are complementary measures rather than a single binary cutoff. [fact; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf]

The occupations most exposed to automation are clerical, transcription, translation, and other structured cognitive roles, while the occupations most exposed to augmentation are specialist roles that still depend on judgment, accountability, and mixed task portfolios. [fact; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]

For workforce planning, the practical implication is to redesign roles and training around task mix: automate repetitive sub-tasks, reskill workers leaving clerical pipelines, and deliberately strengthen AI literacy in occupations that remain human-led but AI-assisted. [inference; source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf]

Key Findings

The accessible Srinivasan research measures automation and augmentation as different constructs, so it does not present a formal crossover rule in which a higher automation score than augmentation score automatically means total occupational replacement. ([fact]; medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
The working paper finds that occupations in the top quartile of automation exposure experienced a 17% decline in job postings per firm per quarter after ChatGPT, while occupations in the top quartile of augmentation exposure experienced a 22% increase. ([fact]; medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
The later Harvard Business School Working Knowledge summary reports updated effects, including a 13% decline for structured repetitive occupations and 20% growth for more analytical, technical, or creative occupations. ([fact]; medium confidence; source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs)
The most automation-exposed occupations are concentrated in clerical and codifiable language work, including correspondence clerks, interpreters and translators, court clerks, medical transcriptionists, telemarketers, typists, and payroll clerks. ([fact]; medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
The most augmentation-exposed occupations are mixed-task specialist roles such as clinical neuropsychologists, medical dosimetrists, agricultural engineers, cartographers, microbiologists, and mediators, where AI can compress sub-tasks but not displace the need for human judgment or responsibility. ([fact]; medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs)
The study supports a role-redesign thesis more than an occupation-extinction thesis, because its outcome variables are posting volume and skill composition rather than observed full elimination of whole job families. ([inference]; medium confidence; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf)
Srinivasan's results align with broader labor-market frameworks from the WEF and McKinsey, both of which also place the highest near-term risk on clerical or routine support work and place the strongest augmentation effects in higher-judgment knowledge work. ([inference]; medium confidence; source: https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier)

Evidence Map

Claim	Source	Confidence	Notes
[fact] Automation and augmentation are defined through separate indices, not a published crossover rule.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	score construction
[fact] Top automation quartile postings fall 17% while top augmentation quartile postings rise 22% in the accessible paper version.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	Table 7
[fact] Updated Harvard Business School summary reports 13% decline and 20% growth.	https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs	medium	later sample window
[fact] Top automation occupations cluster in clerical and codifiable language work.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	Table 3
[fact] Top augmentation occupations cluster in mixed-task specialist roles.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs	medium	Table 3 plus article examples
[inference] The evidence supports role redesign and contraction pressure more strongly than total occupational disappearance.	https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf	medium	postings and skills, not headcount exits
[inference] WEF and McKinsey point in the same direction on routine clerical decline and judgment-rich augmentation.	https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier	medium	directional triangulation

Assumptions

The February 2026 Harvard Business School Working Knowledge article reflects a later revision of the same research program rather than a different model specification, because it cites the same authors, paper title, and qualitative conclusions while updating the sample window and headline percentages. [assumption; source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]
The original infographic likely visualizes a larger set of occupations than the paper's published top-10 table, so this item relies on the table and article examples rather than claiming a complete reconstruction of every plotted point. [assumption; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]

Analysis

The paper traces how generative AI changes labor demand across occupations by linking task composition and skill mix to posting changes. [inference; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf]

That distinction matters because a firm can automate large parts of a role without making the role disappear, especially where non-automatable decision rights, interpersonal accountability, or physical-world execution remain attached to the job. [inference; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]

The best-supported strategic response is a portfolio approach that separates automatable tasks, non-automatable judgment tasks, and new AI-coordination tasks inside each occupation. [inference; source: https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf]

That interpretation also explains why Srinivasan aligns with WEF and McKinsey on clerical decline and high-skill augmentation without collapsing all three frameworks into the same claim, since Srinivasan is grounded in postings and task mix while the others emphasize survey expectations and macro transitions. [inference; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www3.weforum.org/docs/WEF_Future_of_Jobs_2023.pdf; https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america]

Risks, Gaps, and Uncertainties

The accessible paper and the later Harvard Business School summary expose different sample end dates and different effect sizes, so the exact percentages should be treated as version-specific rather than timeless constants. [fact; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]
The accessible sources do not provide a full industry-by-role coordinate dump for the infographic, so the occupation mapping here is strongest for the published top-ranked roles and weaker for exhaustive quadrant reconstruction. [fact; source: https://www.hbs.edu/ris/Publication%20Files/25-039_05fbec84-1f23-459b-8410-e3cd7ab6c88a.pdf; https://www.library.hbs.edu/working-knowledge/enhance-or-eliminate-how-ai-will-likely-change-these-jobs]
The McKinsey comparison is useful for directional triangulation but less precise than the Harvard Business School and WEF evidence in this item because this synthesis relies on McKinsey's official summary framing rather than line-by-line extraction of the underlying report text. [inference; source: https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america; https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier]

Open Questions

How far do the updated 2025 and 2026 versions of the Srinivasan research move the occupation rankings once more post-ChatGPT hiring data is included?
Can the full occupation-by-score dataset behind the Harvard Business School visualization be recovered from a public appendix or data release?
Which entry-level pathways are most vulnerable when postings fall in automation-prone occupations before unemployment visibly rises?