Biochemists and Biophysicists

19-1021.00

Median wage $127,410/yr33,830 employed (US)Rank #540 of 923 scored · top 59% by substitution

Study the chemical composition or physical principles of living cells and organisms, their electrical and mechanical energy, and related phenomena. May conduct research to further understanding of the complex chemical combinations and reactions involved in metabolism, reproduction, growth, and heredity. May determine the effects of foods, drugs, serums, hormones, and other substances on tissues and vital processes of living organisms.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution25

Exposure23

Augmentation70

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

24 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

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%22

panel mean rating 1.9/5 → substitution pressure 22/100

Technical feasibility todayw 20%23

panel mean rating 1.9/5 → substitution pressure 23/100

Cost vs. human wagew 15%22

panel mean rating 1.9/5 → substitution pressure 22/100

Adoption barriersw 20%inverted — strong barriers lower the score32

panel mean rating 3.7/5 (barrier strength) → substitution pressure 32/100

Sector adoption velocityw 10%32

panel mean rating 2.3/5 → substitution pressure 32/100

Task breakdown (24 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.

Determine the three-dimensional structure of biological macromolecules.

CI 49–69 · exposure 47 · augmentation 88 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	5/5	Adoption in structural biology and pharmaceutical research has been rapid and deep since AlphaFold's 2020 release. Major pharma, biotech, and academic labs have integrated AI structure prediction into routine workflows, making it one of the fastest-adopted AI tools in life sciences.
Sector adoption velocity	claude-sonnet-5	4/5	Structural biology and biochemistry research fields have rapidly and deeply adopted AI structure-prediction tools like AlphaFold, reflecting fast uptake typical of computational science domains.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI structure prediction substantially augments researcher productivity by providing initial models in seconds, allowing researchers to focus on validation, refinement, and mechanistic interpretation rather than experimental structure determination. The human expert remains in the loop for critical decisions.
Augmentation potential	claude-sonnet-5	5/5	AI dramatically accelerates hypothesis generation and structure prediction, letting biochemists focus experimental effort on validation and refinement, transforming productivity while keeping humans in the loop.
Task automatability	claude-haiku-4-5-20251001	3/5	AI systems (e.g., AlphaFold, OmegaFold) can now predict protein structures from sequences with high accuracy, dramatically reducing experimental time. However, the full task—including validation, refinement, interpretation, and handling novel or complex systems—still requires substantial human expertise and manual intervention, falling short of the ≥50% time-saving bar for complete end-to-end automation.
Task automatability	claude-sonnet-5	2/5	AI tools like AlphaFold predict structures from sequence with high accuracy, but experimental determination (crystallography, cryo-EM) and validation still require substantial wet-lab and human expert work that AI cannot fully replace end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	2/5	Few regulatory or licensing barriers exist for using AI structure prediction; it is a tool adopted freely by researchers. The main friction is organizational (researcher familiarity, integration into pipelines) and scientific (need for human validation of results), not legal or authorization-based.
Adoption barriers	claude-sonnet-5	2/5	No formal licensing requirement for structure determination itself, though publication and scientific validation norms create moderate friction requiring expert oversight and experimental confirmation.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	Computational structure prediction costs pennies to dollars per protein, orders of magnitude cheaper than experimental determination (X-ray crystallography, cryo-EM) which can cost thousands to tens of thousands. Even with researcher time for validation and interpretation, the AI cost per task-equivalent is substantially lower than the alternative.
Cost vs. human wage	claude-sonnet-5	3/5	Prediction via AI is extremely cheap compared to experimental determination, but many structures still require costly experimental validation, so overall cost savings are task-dependent and not uniformly an order of magnitude cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	4/5	AlphaFold and similar tools are deployed in production at scale (DeepMind, ESMFold, community servers) and deliver reliable predictions for most protein structures. Limitations remain for transmembrane proteins, complexes, and low-confidence regions, but the core task of structure prediction from sequence is demonstrably performed reliably by mature products in real research organizations.
Technical feasibility today	claude-sonnet-5	3/5	AlphaFold and similar tools are deployed and widely used in production research settings for structure prediction, but experimental structure determination still relies on lab instrumentation and human interpretation, limiting full-task reliability.

Write grant proposals to obtain funding for research.

CI 31–54 · exposure 38 · augmentation 88 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Research institutions are beginning to experiment with AI writing tools for grant drafting, and adoption of general research writing assistants is increasing. However, uptake remains exploratory rather than systematic, and many researchers and reviewers remain skeptical of AI-generated proposals.
Sector adoption velocity	claude-sonnet-5	3/5	Academic research is a knowledge-work sector with growing AI tool adoption for writing assistance, but grant writing specifically is conservative due to funder scrutiny and career stakes, keeping adoption at pilot/informal-use levels rather than deep integration.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI significantly augments grant writing by accelerating literature synthesis, improving drafting clarity, organizing arguments, and reducing time spent on formatting and prose editing. Researchers retain full control and judgment while gaining substantial productivity lift on the mechanical aspects of proposal composition.
Augmentation potential	claude-sonnet-5	5/5	AI writing assistants substantially speed up drafting, editing, formatting, and literature summarization for grant proposals, letting biochemists focus scientific reasoning while AI handles prose generation and revision.
Task automatability	claude-haiku-4-5-20251001	2/5	AI can draft sections of grant proposals (literature review, methods) and improve clarity, but cannot replace the human researcher's original scientific vision, innovation framing, or the credibility signal that the PI's voice provides. End-to-end automation without significant human oversight would result in loss of scientific authenticity and likely rejection.
Task automatability	claude-sonnet-5	3/5	AI can draft significant portions of grant text (background, specific aims, methods framing) given researcher input, saving substantial time, but crafting a fundable proposal requires original scientific insight, novel hypotheses, and strategic framing that still needs heavy human authorship and revision.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Grant proposals must be signed by and represent the intellectual contribution of the human researcher; funding agencies expect human authorship and accountability. Institutional review, PI certification, and funder skepticism of AI-authored proposals create strong organizational and reputational barriers to full automation.
Adoption barriers	claude-sonnet-5	3/5	No licensing requirement bars AI use, but funding agencies require PI accountability, scientific integrity, and often disclosure of AI use in preparation, creating moderate institutional and reputational friction.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	Using AI for grant drafting (inference + integration) costs pennies per proposal compared to dozens of hours of researcher time or professional grant writing services. The cost differential is highly favorable, though this does not translate to replacement since human judgment remains essential.
Cost vs. human wage	claude-sonnet-5	3/5	AI drafting tools are cheap relative to researcher time, but the scientist must still invest substantial time reviewing, correcting technical content, and ensuring scientific rigor, so net cost savings are moderate rather than order-of-magnitude.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI writing assistants (ChatGPT, Claude) can generate proposal text, no deployed system reliably produces fundable grant proposals autonomously. Existing tools are used as drafting aids, not as end-to-end systems. Grant success depends on scientific novelty and institutional reputation, which AI cannot establish.
Technical feasibility today	claude-sonnet-5	3/5	LLM-based writing assistants (ChatGPT, Claude, specialized grant-writing tools) are widely used by researchers today to draft sections and improve language, but no product reliably produces fundable, technically sound proposals without extensive scientist oversight.

Prepare reports or recommendations, based upon research outcomes.

CI 30–55 · exposure 42 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption of AI for report writing in biochemistry and biophysics remains limited to exploratory pilots in large research institutions. Most labs rely on traditional authorship by humans; regulatory and funding bodies (NIH, journals) have not yet normalized AI-generated research reports.
Sector adoption velocity	claude-sonnet-5	3/5	R&D and biotech/pharma sectors are adopting AI writing and analysis tools at a moderate pace, with pilots and partial integration common but full automated reporting still rare.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI can meaningfully assist by drafting initial report sections, organizing results, suggesting visualizations, and flagging potential gaps in reasoning—significantly raising a biochemist's drafting speed while the expert retains final judgment over all recommendations.
Augmentation potential	claude-sonnet-5	4/5	AI significantly speeds up drafting, summarizing data trends, and formatting recommendations, meaningfully boosting productivity while the biochemist retains responsibility for interpretation and accuracy.
Task automatability	claude-haiku-4-5-20251001	2/5	AI can draft sections of reports (methods, summary statistics) and suggest organizational structures, but synthesizing complex research outcomes into sound recommendations requires domain expertise, judgment about significance, and often subjective interpretation of results. This typically saves <50% time because the human must substantially rework and validate the core intellectual content.
Task automatability	claude-sonnet-5	3/5	AI can draft report sections, summarize data, and generate recommendation language from structured research outputs, but synthesizing novel scientific conclusions and validating them requires expert judgment AI cannot fully replace end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Publication and research integrity standards require the responsible researcher (typically a licensed or credentialed scientist) to personally vouch for conclusions and recommendations. Liability for incorrect recommendations, regulatory/funder expectations of human expert judgment, and institutional norms create strong barriers to full automation.
Adoption barriers	claude-sonnet-5	2/5	No formal licensing requirement for report authorship, but scientific integrity norms, publication/peer review expectations, and institutional accountability create moderate friction against pure AI-generated conclusions.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI inference and integration for report generation is inexpensive, but the task requires significant expert human review and iteration to ensure scientific accuracy and validity. The total cost remains comparable to a human biochemist writing the report from scratch, given necessary oversight.
Cost vs. human wage	claude-sonnet-5	3/5	AI drafting reduces some writing time cheaply, but the need for expert scientist review and validation of technical accuracy keeps overall cost comparable to human-led effort rather than dramatically cheaper.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Large language models can generate report drafts and summaries from research data; some research institutions use AI for literature synthesis and result visualization. However, no deployed product reliably handles the full end-to-end task of interpreting novel biochemical findings and formulating defensible recommendations without material human revision.
Technical feasibility today	claude-sonnet-5	3/5	LLM-based writing assistants and lab-notebook/report tools are deployed in scientific settings today, but they require heavy human editing and fact-checking, limiting reliability for final scientific reports.

Share research findings by writing scientific articles or by making presentations at scientific conferences.

CI 28–50 · exposure 38 · augmentation 75 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Academic and research institutions are experimenting with AI writing aids and generating interest in automation, but actual adoption remains in pilot and informal-assistance phases. Production-level replacement is rare; most usage is augmentative and institution-dependent.
Sector adoption velocity	claude-sonnet-5	3/5	Academic and scientific research is a knowledge-intensive sector with growing AI tool adoption for writing assistance, but formal publication and presentation practices remain conservative and slow to change norms.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI excels at drafting abstracts, outlining arguments, generating first-pass text, organizing citations, and preparing presentation slides. Biochemists can substantially accelerate initial composition and revision cycles, though they must retain full expert control over content accuracy and scientific claims.
Augmentation potential	claude-sonnet-5	4/5	AI substantially helps with drafting, editing, literature summarization, and creating presentation materials, meaningfully speeding up the writing and dissemination process while the scientist retains control over content and conclusions.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist in drafting sections and organizing content, writing publication-quality scientific articles requires novel hypothesis formation, interpretation of complex experimental data, and critical judgment that AI cannot reliably perform end-to-end. Presentations demand similar novelty and authority that current systems cannot independently deliver at the quality and originality expected in peer-reviewed venues.
Task automatability	claude-sonnet-5	3/5	AI can draft manuscript sections, format citations, and generate slide decks, but synthesizing novel findings, ensuring scientific accuracy, and framing contributions for peer review still require substantial human expertise and revision.dusk
Adoption barriers	claude-haiku-4-5-20251001	4/5	Scientific articles must bear the author's reputation and institutional accountability; liability and error-cost asymmetry are severe (false claims damage careers and science). Peer review and journal gatekeeping inherently require human expert judgment, and authorship conventions legally and professionally bind the human scientist to the work's integrity.
Adoption barriers	claude-sonnet-5	3/5	No licensing requirement blocks AI use, but journal policies on AI-generated content, authorship norms, and concerns about scientific integrity create meaningful friction.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Current AI services (large language models, writing assistants) cost relatively little per output, but the overhead of expert review, fact-checking, and rewriting to meet publication standards means the effective cost-per-usable-output remains high compared to a human scientist directly writing the material.
Cost vs. human wage	claude-sonnet-5	3/5	AI writing assistance is cheap relative to researcher time spent drafting, but the need for expert review, fact-checking, and revision keeps overall costs comparable to traditional writing workflows.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI writing tools exist but produce generic, error-prone outputs that require extensive human revision for scientific accuracy and novelty. No deployed product reliably writes complete, citable scientific articles or presentation slides without substantial human expert oversight and rewriting.
Technical feasibility today	claude-sonnet-5	3/5	Tools like ChatGPT, Grammarly, and AI writing assistants are widely used by researchers for drafting and editing, but no product reliably produces publication-ready scientific articles or conference talks without heavy human oversight.

Study the mutations in organisms that lead to cancer or other diseases.

CI 28–47 · exposure 30 · augmentation 88 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	4/5	Genomics and bioinformatics have rapidly adopted AI tools; variant calling, annotation, and pathogenicity prediction are now routine in research and clinical labs, with machine learning models widely integrated into analysis pipelines.
Sector adoption velocity	claude-sonnet-5	3/5	Genomics and computational biology have seen notable AI tool adoption (e.g., AlphaFold, variant annotation pipelines) but wet-lab and hypothesis-driven mutation research remains a mixed adoption environment with pilots more common than full production integration.
Augmentation potential	claude-haiku-4-5-20251001	5/5	AI substantially augments this task by rapidly processing large genomic datasets, prioritizing mutations of interest, and generating evidence summaries, allowing biochemists to focus on mechanistic investigation and clinical interpretation rather than manual data screening.
Augmentation potential	claude-sonnet-5	4/5	AI substantially augments this work by accelerating literature review, variant interpretation, structural prediction, and generating hypotheses for further experimental testing, while scientists remain essential for validation and interpretation.
Task automatability	claude-haiku-4-5-20251001	2/5	AI can help identify mutation patterns and perform bioinformatics analysis on sequence data, but the task requires hypothesis formation, experimental design, and interpretation of complex biological mechanisms that cannot yet be fully automated end-to-end at the required quality level.
Task automatability	claude-sonnet-5	2/5	AI can assist in analyzing genomic data and literature but the core experimental design, wet-lab validation, and hypothesis generation for mutation-disease causation still require substantial human expertise and lab work that current AI cannot fully replace.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory oversight (FDA approval for clinical-grade variant interpretation), publication and institutional review standards, and the requirement for expert human judgment in evaluating mechanistic significance and novelty create substantial barriers to full automation.
Adoption barriers	claude-sonnet-5	3/5	No strict licensing requirement for research itself, but findings on disease-causing mutations typically require peer review, institutional oversight (IRB/ethics), and expert validation before being trusted or acted upon clinically.
Cost vs. human wage	claude-haiku-4-5-20251001	4/5	Computational analysis of mutations is increasingly cheap compared to manual labor, though integration into existing workflows and expert oversight still require human effort; the per-analysis cost of AI-driven sequence analysis is a small fraction of a biochemist's hourly rate.
Cost vs. human wage	claude-sonnet-5	2/5	AI tools reduce some computational and literature-review costs, but the overall research process (experiments, validation, interpretation) still requires expensive skilled labor and lab infrastructure, keeping AI's cost advantage limited to sub-components.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed bioinformatics tools and AI-assisted variant analysis systems exist and are widely used in research and clinical settings, but they typically function as assistive rather than autonomous systems, requiring substantial expert interpretation and validation.
Technical feasibility today	claude-sonnet-5	2/5	Deployed bioinformatics tools (variant callers, pathogenicity predictors like AlphaMissense) exist and are used in production, but the broader task of studying mutation mechanisms and disease causation still requires human-led research programs with only partial AI support.

Investigate the nature, composition, or expression of genes or research how genetic engineering can impact these processes.

CI 28–37 · exposure 30 · augmentation 75 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	3/5	Academic and pharmaceutical sectors are integrating AI-assisted computational tools (structure prediction, library design) into workflows, but actual displacement of core genetic investigation remains limited; most adoption is augmentative rather than substitutive, reflecting the complexity and regulatory constraints of the field.
Sector adoption velocity	claude-sonnet-5	3/5	Life sciences and biotech have adopted AI tools rapidly for specific tasks (protein folding, sequence design) but broader genetic research automation remains at pilot/tool-assisted stage rather than deep production integration.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI substantially augments genetic research through rapid sequence annotation, structure prediction, and hypothesis generation from large datasets, enabling researchers to focus on experimental design and validation while AI handles computational screening and literature synthesis.
Augmentation potential	claude-sonnet-5	4/5	AI substantially augments this work through tools like AlphaFold, CRISPR design software, literature mining, and bioinformatics pipelines, significantly accelerating hypothesis generation and data analysis while humans remain central to experimental design and interpretation.
Task automatability	claude-haiku-4-5-20251001	2/5	Current AI can assist with literature review, sequence analysis, and computational prediction of genetic interactions, but cannot independently design, conduct, or interpret wet-lab experiments or navigate the complex exploratory judgment required for novel genetic investigations at equal quality and time savings.
Task automatability	claude-sonnet-5	2/5	This is open-ended experimental and hypothesis-driven research involving wet-lab work, novel discovery, and physical genetic manipulation that current AI cannot execute end-to-end; AI can assist with analysis and literature synthesis but not the core investigative process.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory oversight (FDA, IRB, institutional biosafety committees) mandates human researcher accountability for genetic engineering work; liability for unintended genetic modifications is asymmetrically borne by the organization; and intellectual property and publication norms require human scientific judgment and sign-off.
Adoption barriers	claude-sonnet-5	3/5	While no strict licensing mandates a human perform genetic research, biosafety regulations, IRB/IACUC oversight, and publication/peer-review norms create moderate institutional and regulatory friction around genetic engineering research.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI tools reduce computational overhead and accelerate screening, but the specialized infrastructure, wet-lab equipment, and senior researcher time required for actual genetic investigation remain far more expensive than AI inference; cost parity has not yet been achieved.
Cost vs. human wage	claude-sonnet-5	2/5	AI computational tools are cheap for specific analyses, but the overall research task requires expensive lab infrastructure, reagents, and expert oversight, keeping total cost comparable to or dependent on human expertise.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	Deployed AI tools (AlphaFold, sequence aligners, computational biology platforms) reliably perform narrow subtasks like structure prediction and homology searches, but no end-to-end system independently investigates novel gene expression or validates genetic engineering impacts without substantial human direction and wet-lab validation.
Technical feasibility today	claude-sonnet-5	2/5	AI tools (e.g., AlphaFold, sequence analysis models, genomics platforms) are deployed for narrow sub-tasks like structure prediction or variant annotation, but no product autonomously conducts genetic engineering research or novel gene-function investigation in production.

Design or build laboratory equipment needed for special research projects.

CI 7–52 · exposure 33 · augmentation 63 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption of AI for equipment design in biochemistry and biophysics remains limited; most research groups still rely on traditional design, vendor equipment, and experienced technicians. While digitization in research is increasing, actual production deployment of AI-driven equipment design in this sector is still in pilot phases.
Sector adoption velocity	claude-sonnet-5	2/5	Academic and research lab environments adopt AI slowly for physical equipment design, though some CAD/generative design tools are trickling in for narrow use cases.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI can significantly augment researcher productivity by generating design options, automating CAD drafting, performing simulations, and suggesting component selections. This keeps the human researcher in the loop for validation and iteration, substantially speeding the design phase without removing human judgment.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with CAD modeling, simulation, and literature-based design ideation, improving efficiency in early planning stages of equipment design.
Task automatability	claude-haiku-4-5-20251001	4/5	AI can assist in equipment design (CAD generation, component selection, optimization algorithms) and help generate specifications or procurement lists, achieving significant time savings. However, the task requires empirical validation, physical prototyping, and domain expertise that currently requires human iteration and testing, preventing full end-to-end automation.
Task automatability	claude-sonnet-5	1/5	Designing/building novel lab equipment requires physical engineering, hands-on prototyping, and creative problem-solving that current AI cannot execute end-to-end without extensive human physical labor and judgment.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Significant barriers exist: regulatory requirements for laboratory equipment, safety certification needs, organizational friction around adopting AI for design in research contexts, and the need for a qualified human (often the researcher) to sign off on specifications and validate performance before construction.
Adoption barriers	claude-sonnet-5	3/5	No licensing requirement per se, but safety, custom engineering needs, and reliance on specialized technical skill create moderate organizational friction against automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	While AI design assistance reduces labor hours, the equipment building phase remains largely manual and requires skilled technicians. The cost of AI tools, integration, and oversight adds up to remain comparable to or exceeds the value of human labor saved, particularly for custom, low-volume research equipment.
Cost vs. human wage	claude-sonnet-5	1/5	Physical fabrication, sourcing parts, and hands-on assembly still require skilled human labor and machining, so AI does not reduce cost per equivalent output today.
Technical feasibility today	claude-haiku-4-5-20251001	3/5	AI tools for design (generative CAD, simulation software) exist in production, but they require substantial human oversight and validation for novel research equipment. Current systems can draft designs and suggest components reliably, but cannot independently verify performance or handle the iterative refinement needed for specialized equipment.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously designs and builds custom laboratory apparatus; this remains a manual, expert-driven engineering task with only isolated CAD or simulation assistance.

Study spatial configurations of submicroscopic molecules, such as proteins, using x-rays or electron microscopes.

CI 25–32 · exposure 25 · augmentation 88 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Biochemistry and biophysics are relatively laggard in AI automation adoption for experimental tasks; while computational tools are used, active research labs still rely on human-directed experimentation, and automation is limited to post-hoc analysis rather than end-to-end experimental autonomy.
Sector adoption velocity	claude-sonnet-5	3/5	Computational biology/pharma sectors are adopting AI structure-prediction tools rapidly, but the physical experimental side of structural biology remains slower to be transformed by AI compared to purely digital knowledge work.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI substantially augments human productivity in this domain: AlphaFold and similar tools accelerate structure prediction, automated image analysis speeds up microscopy data interpretation, and machine learning assists in hypothesis generation—all while the biochemist remains central to experimental direction and validation.
Augmentation potential	claude-sonnet-5	5/5	Tools like AlphaFold, RoseTTAFold, and AI-assisted cryo-EM image processing dramatically speed up structure determination and hypothesis generation, transforming researcher productivity while humans still design experiments and validate results.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist in image analysis and molecular structure prediction (e.g., AlphaFold for protein folding), the end-to-end task requires experimental design, data collection using specialized equipment, interpretation of results, and validation—steps that demand human expertise and cannot be reliably automated to 50% time savings today.
Task automatability	claude-sonnet-5	2/5	AI (e.g., AlphaFold-class tools) can predict structures computationally, but the actual acquisition and interpretation of X-ray crystallography or cryo-EM data from physical samples requires lab work, instrument operation, and validation that current AI cannot fully replace.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Significant barriers exist: regulatory requirements for laboratory safety and equipment operation, the need for specialized credentials and training, liability concerns with incorrect structural interpretations, and the requirement for human judgment in experimental design and validation of results.
Adoption barriers	claude-sonnet-5	3/5	No licensing barrier per se, but this task requires specialized lab access, physical instrumentation, and domain expertise, creating practical barriers to full automation even though no formal sign-off requirement exists.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The task requires expensive specialized equipment (X-ray crystallography, electron microscopes) and skilled operators that dominate cost; AI inference alone is cheap but cannot replace the equipment and hands-on experimental work, making the total cost structure heavily dependent on human labor and capital.
Cost vs. human wage	claude-sonnet-5	2/5	Experimental structural biology requires expensive instrumentation, sample prep, and skilled technicians; AI can reduce some modeling costs but does not eliminate the substantial fixed costs of running X-ray/electron microscopy studies, so overall cost savings versus human labor are modest.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI tools exist for protein structure prediction and image analysis, but they are assistive rather than end-to-end solutions. No deployed system reliably performs the full experimental pipeline from sample preparation through microscopy operation to validated structure determination without substantial human oversight.
Technical feasibility today	claude-sonnet-5	2/5	Structure-prediction tools are deployed and widely used, but the experimental task described (collecting/analyzing X-ray or electron microscope data on physical molecules) still relies heavily on human-operated instrumentation and expert interpretation, with AI serving as a complement rather than an autonomous performer.

Examine the molecular or chemical aspects of immune system functioning.

CI 25–32 · exposure 25 · augmentation 63 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	While biopharmaceutical and academic sectors use AI for drug screening and data analysis, actual autonomous examination of immune system functioning remains limited; adoption is primarily in screening and optimization roles rather than primary investigation, reflecting slow displacement of core research tasks.
Sector adoption velocity	claude-sonnet-5	3/5	Life sciences and biomedical research have growing AI tool adoption (protein structure prediction, literature synthesis, data analysis) but many labs still use these as pilot/augmentation tools rather than deep production-scale automation.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI tools meaningfully augment biochemists by accelerating literature review, suggesting molecular candidates, visualizing complex structures, and analyzing high-throughput data, but the human expert remains essential for experimental design, interpretation, and novel discovery.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids literature review, hypothesis generation, data analysis, and structural modeling, meaningfully increasing researcher productivity while humans retain control of experimental design and interpretation.
Task automatability	claude-haiku-4-5-20251001	2/5	AI can assist with analyzing existing immunological data, simulating molecular interactions, and literature synthesis, but generating novel immunological insights requires deep domain expertise, experimental validation design, and contextual judgment that current systems cannot reliably perform end-to-end without substantial human oversight.
Task automatability	claude-sonnet-5	2/5	This task involves designing experiments, generating novel hypotheses, and interpreting complex biological data requiring physical lab work and deep domain judgment that current AI cannot perform end-to-end.assistance is significant but full automation is far from a 50% time-saving threshold across the whole task.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory oversight (FDA, institutional review boards), liability for incorrect immunological findings, requirement for qualified personnel to design and validate experiments, and institutional trust in human expertise create substantial adoption barriers; autonomous AI-driven immune research faces legal and ethical constraints.
Adoption barriers	claude-sonnet-5	3/5	No formal licensing requires a human to do this specific research task, but publication norms, scientific accountability, peer review, and grant-funded PI oversight create meaningful institutional friction against full AI substitution.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	Current AI tools for molecular analysis and simulation require significant computational infrastructure, integration with specialized pipelines, and expert human interpretation; the combined cost approaches or exceeds that of a trained biochemist for equivalent novel insights.
Cost vs. human wage	claude-sonnet-5	2/5	Wet-lab experimentation, reagents, and expert interpretation dominate cost; AI reduces some analysis/literature time but does not replace the bulk of costly experimental and interpretive labor, so overall cost savings are modest.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Deployed AI tools can support molecular visualization, data analysis, and hypothesis generation from published literature, but no product reliably performs independent examination of immune system functioning at the depth and accuracy required in research or diagnostic settings; most capabilities remain in the research or narrow-application phase.
Technical feasibility today	claude-sonnet-5	2/5	AI tools (AlphaFold, literature-mining assistants) support narrow sub-components like structure prediction or literature review, but no deployed product autonomously examines molecular/chemical immune mechanisms as a complete research task in production.

Isolate, analyze, or synthesize vitamins, hormones, allergens, minerals, or enzymes and determine their effects on body functions.

CI 25–30 · exposure 25 · augmentation 63 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Biotech and pharma sectors use AI for in silico screening and data mining, but adoption of fully automated compound isolation and synthesis remains limited to specialized robotics labs. The field still heavily relies on traditional bench experimentation and human expertise, with AI in supporting rather than leading roles.
Sector adoption velocity	claude-sonnet-5	2/5	Biotech and pharma R&D are adopting AI for specific tasks (molecule design, literature synthesis) but wet-lab biochemistry work remains largely manual with slow, uneven AI integration.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI meaningfully assists biochemists by predicting protein structures, suggesting reaction pathways, accelerating literature review, and analyzing spectroscopic data, allowing researchers to focus on hypothesis generation and experimental design. However, the impact is primarily on the analytical and planning phases rather than the hands-on synthesis work itself.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids literature review, data analysis, structure prediction, and experimental design, meaningfully speeding up parts of this research process even though the physical work remains human-led.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with data analysis and literature synthesis for understanding biochemical mechanisms, the core task requires hands-on laboratory work to isolate and synthesize compounds—centrifugation, chromatography, spectroscopy—which cannot be automated by current AI systems. AI cannot physically manipulate samples or operate equipment without robotic systems, which are rare in this context.
Task automatability	claude-sonnet-5	2/5	This involves substantial wet-lab work (isolation, synthesis, physiological assays) that requires physical manipulation and instrumentation AI cannot perform end-to-end; only data analysis and literature portions are automatable today.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Regulatory compliance (FDA, EMA) requires validated experimental protocols and human expert sign-off on clinical or pharmaceutical applications. Institutional review boards, GLP/GMP standards, and liability frameworks mandate human responsibility for biochemical analysis, particularly for compounds intended for therapeutic or food use.
Adoption barriers	claude-sonnet-5	3/5	No licensure requirement for the analysis itself, but biosafety, regulatory compliance (FDA/IRB), and physical lab safety protocols create moderate friction against full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI-assisted bioinformatic analysis is cheap relative to computational time, but the dominant cost is the biochemist's labor for experimental design, execution, and interpretation. Current AI cannot eliminate the need for skilled human operators, so all-in costs remain high relative to labor savings.
Cost vs. human wage	claude-sonnet-5	2/5	Physical lab work, reagents, and specialized equipment dominate costs; AI can cut some analysis/design time but the overall task still requires expensive human-operated lab infrastructure, so savings are partial.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed AI product performs end-to-end isolation, analysis, or synthesis of biochemical compounds independently. AI tools exist for predictive modeling and structure analysis, but they complement rather than replace wet-lab experimentation and remain limited to narrow domains without human expert validation.
Technical feasibility today	claude-sonnet-5	2/5	AI tools (e.g., protein structure prediction, cheminformatics, lab automation robots) exist for narrow sub-steps, but no deployed product performs the full isolate-analyze-synthesize-determine-effects workflow reliably in production.

Research how characteristics of plants or animals are carried through successive generations.

CI 25–30 · exposure 25 · augmentation 75 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Academic and biotech research sectors show cautious, slow adoption of AI for autonomous research. While AI is widely used for data analysis and literature review, independent research design and execution remain almost entirely human-driven. Pilots are emerging but production-scale autonomous research is rare.
Sector adoption velocity	claude-sonnet-5	2/5	Academic and biotech research sectors are adopting AI tools steadily but cautiously, with pilots for data analysis common but full task automation rare given the specialized, hypothesis-driven nature of genetics research.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI substantially augments human geneticists through sequence analysis, literature mining, statistical modeling of inheritance patterns, and hypothesis suggestion from large datasets. These tools meaningfully accelerate the research workflow while the biochemist retains experimental design, biological judgment, and interpretation authority.
Augmentation potential	claude-sonnet-5	4/5	AI substantially augments this task through tools like AlphaFold, genomic analysis pipelines, and literature-mining assistants that accelerate research while scientists maintain interpretive control.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can analyze existing genetic sequences and predict inheritance patterns from data, the task requires original research involving experimental design, hypothesis formation, and interpretation of novel biological phenomena. Current AI systems lack the autonomous capability to design breeding experiments, collect organisms, observe multigenerational traits, and synthesize novel findings at the level required for independent research.
Task automatability	claude-sonnet-5	2/5	This is open-ended scientific research involving experimental design, wet-lab work, and hypothesis generation about heredity mechanisms; AI can assist literature review and data analysis but cannot independently conduct the biological research end-to-end.dream A large fraction of the task—physical experimentation, novel hypothesis formation—remains human-driven.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Significant barriers exist: research integrity standards require human researchers to take responsibility for experimental design and interpretation; institutional review boards oversee animal/plant research; publication and peer review systems expect human authorship and accountability. Legal liability for novel organisms or failed experiments falls on licensed researchers.
Adoption barriers	claude-sonnet-5	3/5	No licensing requirement specifically bars AI from research, but scientific credibility, peer review, and institutional trust in human-authored research create moderate friction against full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The computational cost of AI genomic analysis is low, but the task encompasses experimental design, organism maintenance, and novel hypothesis generation—domains where human biochemists remain significantly cheaper than building autonomous experimental systems. Integration and validation costs remain high relative to expert human labor.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cheaply handle computational subtasks like sequence analysis, but the overall research task still requires expensive human expertise, lab infrastructure, and oversight, keeping costs comparable to human-only research.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI tools can assist in sequence analysis and pattern recognition within genetic data, but no deployed product performs end-to-end genetics research autonomously. Existing systems require substantial human expertise to design experiments, interpret biological context, and validate findings against real-world observations.
Technical feasibility today	claude-sonnet-5	2/5	Deployed AI tools help with genomic data analysis, literature synthesis, and sequence prediction, but no product autonomously conducts inheritance research studies in production at scale.

Research the chemical effects of substances, such as drugs, serums, hormones, or food, on tissues or vital processes.

CI 25–25 · exposure 25 · augmentation 75 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption of AI in biochemical research is moderate and incremental—primarily in data mining and in silico prediction—but wet-lab and translational research remain heavily manual. Academic and pharmaceutical labs are slowly integrating AI assistants, not replacing researchers or automating the core investigation task.
Sector adoption velocity	claude-sonnet-5	2/5	Life sciences R&D is adopting AI for specific subtasks (target identification, literature review) but wet-lab biochemistry/biophysics research remains slow to digitize compared to information-sector benchmarks.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI substantially augments biochemists by accelerating literature discovery, predicting molecular interactions, analyzing high-dimensional assay data, and suggesting hypotheses; these capabilities meaningfully raise productivity while the researcher remains essential for experimental design, validation, and interpretation of complex biological mechanisms.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids literature review, experimental design suggestions, data analysis, and molecular modeling, meaningfully increasing researcher productivity while humans remain essential for hands-on experimentation and interpretation.
Task automatability	claude-haiku-4-5-20251001	2/5	Current AI can assist with literature review, hypothesis generation, and data analysis of existing chemical/biological data, but cannot independently design experiments, conduct wet-lab work, or synthesize novel findings from raw experimental data in a way that meets the 50% time-saving bar for the full research task. Experimental design and execution remain human-dependent.
Task automatability	claude-sonnet-5	2/5	Wet-lab experimentation, sample handling, and physiological observation are core to this task and cannot be performed by current AI; only literature synthesis and hypothesis generation portions are automatable.", ...},
Adoption barriers	claude-haiku-4-5-20251001	4/5	Research involving drugs, serums, and hormones faces significant regulatory barriers (FDA, IRB approval, Good Laboratory Practice compliance) and requires licensed biochemists to design, execute, and validate experiments. Liability for incorrect chemical effect assessment creates strong organizational friction against full automation.
Adoption barriers	claude-sonnet-5	4/5	Research involving drug and biological effects is subject to regulatory oversight (IRB/IACUC, FDA), safety protocols, and requires credentialed scientists to interpret and validate results, creating strong barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI-assisted tools (cloud compute, software licenses) reduce time on secondary tasks like literature review, but the core experimental work, reagent costs, and skilled labor remain expensive. Full-stack automation savings do not yet offset the high cost of maintaining research infrastructure and expert validation.
Cost vs. human wage	claude-sonnet-5	2/5	Physical experimentation, reagents, and lab infrastructure dominate costs; AI only reduces costs on the analysis/literature-review slice, so overall cost savings versus a human researcher are modest.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI tools like ChemBERTA, molecular docking software, and large language models can support chemical effect prediction and literature synthesis, but no deployed product reliably performs end-to-end biochemical research—evaluating complex tissue responses, interpreting assay results, or validating chemical mechanisms—without substantial human oversight and experimental validation.
Technical feasibility today	claude-sonnet-5	2/5	AI tools (e.g., AlphaFold, literature-mining assistants) support hypothesis generation and data analysis but no deployed product independently conducts substance-tissue effect research end-to-end in production settings.

Develop or execute tests to detect diseases, genetic disorders, or other abnormalities.

CI 25–25 · exposure 25 · augmentation 63 · importance 3.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Adoption of AI in clinical diagnostics is growing but concentrated in imaging and genomic variant calling—not in autonomous test development or execution. Most biochemistry labs remain traditional, with AI supporting analysis rather than driving automation of hands-on testing workflows at scale.
Sector adoption velocity	claude-sonnet-5	2/5	Biomedical research and clinical lab sectors adopt AI tools incrementally for specific analytical tasks, but wet-lab test development remains slow to digitize compared to purely information-based fields.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI augments biochemists by automating routine result interpretation, flagging anomalies, and proposing diagnostic hypotheses from complex datasets, which meaningfully assists the analytical phase. However, the development and execution phases (protocol design, instrument operation, quality assurance) see limited current augmentation.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids literature review, experimental design, data analysis, and pattern detection in genetic/diagnostic data, meaningfully increasing researcher productivity while humans perform and validate the physical assay work.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist in analyzing test results and interpreting data patterns from established assays, developing and executing physical tests—sample preparation, instrument operation, quality control, troubleshooting—requires hands-on laboratory work that current AI cannot perform end-to-end. AI lacks the embodied capability and real-time adaptive judgment needed to replace the full workflow.
Task automatability	claude-sonnet-5	2/5	Test development requires substantial hands-on wet-lab work, hypothesis-driven experimental design, and physical execution that current AI cannot perform end-to-end, though AI can assist with data analysis and design components.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Clinical and diagnostic testing is heavily regulated by agencies like the FDA and CLIA; any automated test system must be validated, approved, and traceable to a licensed professional. Liability for false results, requirement for expert sign-off on novel assays, and organizational validation requirements create strong adoption barriers.
Adoption barriers	claude-sonnet-5	4/5	Diagnostic test development and validation are subject to regulatory oversight (e.g., CLIA, FDA), requiring credentialed professionals to validate and sign off on clinical or genetic testing results.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI inference for result analysis is cheap, but integrating it into full test workflows—including equipment, reagents, sample handling, and validation—remains expensive relative to the biochemist labor it might partially offset. The specialized nature of test development commands high wages, making overall cost parity unlikely.
Cost vs. human wage	claude-sonnet-5	2/5	Physical lab execution still requires human labor, equipment, and oversight; AI tools reduce some analysis time but do not eliminate the dominant cost of skilled personnel and wet-lab infrastructure.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI-based diagnostic tools exist for result interpretation (e.g., imaging analysis, genetic variant classification), but deployed products do not reliably execute the complete test development and execution pipeline autonomously. Production systems remain narrow in scope and typically require human interpretation and method validation.
Technical feasibility today	claude-sonnet-5	2/5	Deployed AI products exist for narrow analytical subtasks (e.g., variant calling, image classification for pathology) but no product autonomously develops and executes full diagnostic test protocols in a lab setting.

Develop methods to process, store, or use foods, drugs, or chemical compounds.

CI 20–30 · exposure 25 · augmentation 75 · importance 3.3/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Pharmaceutical and food-science sectors adopt computational aids slowly; most R&D remains lab-based and hands-on. Adoption of standalone AI agents for method development is nascent, with heavy reliance on human biochemists to design, validate, and oversee all stages.
Sector adoption velocity	claude-sonnet-5	2/5	Pharma/biotech and food science sectors are experimenting with AI in R&D (e.g., drug discovery platforms) but broad production-scale adoption for full method development remains limited and slow-moving due to regulatory and safety constraints.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI significantly augments biochemist productivity through computational modeling of molecular interactions, rapid virtual screening of storage conditions, optimization of process parameters, and literature mining. These tools substantially accelerate hypothesis generation and reduce experimental iteration while the scientist retains final validation responsibility.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids literature synthesis, molecular modeling, data analysis, and hypothesis generation, meaningfully speeding up parts of the method development process even though humans remain essential for lab work and validation.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with computational design of processing methods and storage optimization, the task fundamentally requires hands-on laboratory experimentation, regulatory validation, and iterative physical testing that cannot be fully automated. Current AI cannot independently design novel bioprocesses, run experiments, and validate stability/safety at a fraction of human time.
Task automatability	claude-sonnet-5	2/5	This is a creative R&D task requiring novel experimental design, hypothesis generation, and physical lab validation, which AI cannot yet execute end-to-end despite being able to assist with literature review and hypothesis brainstorming.
Adoption barriers	claude-haiku-4-5-20251001	5/5	FDA, EMA, and comparable regulatory bodies require documented human expertise, validation protocols, and signed sign-off by qualified personnel (Ph.D. chemists, M.D.s) for food and pharmaceutical process approval. Liability, safety requirements, and mandatory regulatory review create hard barriers to full automation.
Adoption barriers	claude-sonnet-5	3/5	While no explicit licensing barrier bars AI use, regulatory requirements (FDA, safety testing) and organizational reliance on credentialed scientists to validate and sign off on new methods create moderate friction.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI tools for computational design and simulation can reduce some costs, but the bulk of method development involves physical experimentation, equipment operation, and regulatory compliance that remain labor-intensive. Overall cost per validated method remains higher than or comparable to employing a biochemist team.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cheaply generate candidate ideas or screen compounds computationally, but the overall method development still requires expensive wet-lab validation and expert oversight, keeping total cost comparable to or only modestly below human-only R&D.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	Some computational tools (molecular modeling, parameter optimization software) are deployed for parts of method development, but no end-to-end AI system reliably generates validated food/drug processing methods. Products operate within narrow scopes and require extensive human oversight; deployment in regulated pharmaceutical/food sectors remains limited.
Technical feasibility today	claude-sonnet-5	2/5	AI tools (e.g., protein/molecule design models, literature mining) exist as research aids but no deployed product independently develops complete processing or formulation methods in production settings.

Study the chemistry of living processes, such as cell development, breathing and digestion, or living energy changes, such as growth, aging, or death.

CI 16–30 · exposure 20 · augmentation 75 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Biochemistry remains a specialized, capital-intensive, bench-focused field with slow digital adoption relative to information work. While leading academic labs use AI for modeling and data analysis, production-scale displacement of the research task itself is minimal; most adoption remains at the tool level, not the autonomous agent level.
Sector adoption velocity	claude-sonnet-5	2/5	Academic and pharmaceutical research settings adopt AI tools for literature synthesis and modeling, but core experimental biochemistry work sees only pilot-level AI integration.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI substantially augments biochemists through structure prediction (AlphaFold), molecular docking, literature synthesis, statistical analysis of complex datasets, and hypothesis prioritization. These tools meaningfully raise productivity on analysis and literature review while keeping the scientist in the loop for experimental design and interpretation.
Augmentation potential	claude-sonnet-5	4/5	AI substantially assists literature review, hypothesis generation, protein structure prediction, and data analysis, meaningfully boosting researcher productivity while humans remain central to experimentation.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can accelerate literature review, data analysis, and hypothesis generation, the core task of studying living chemistry requires designing novel experiments, interpreting ambiguous biological results, and making judgment calls on experimental direction. Current AI cannot autonomously conceive and execute wet-lab experiments or reliably synthesize findings into novel mechanistic understanding at the depth biochemistry demands.
Task automatability	claude-sonnet-5	2/5	This describes broad, open-ended scientific research requiring hypothesis generation, wet-lab experimentation, and physical manipulation of biological systems that current AI cannot perform end-to-end.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Biochemical research faces material barriers to automation: institutional review boards govern novel experiments, safety protocols protect researchers and organisms, peer-review norms require human authorship and accountability, and regulatory pathways (e.g., pharmaceutical preclinical work) often require licensed scientists to sign off on conclusions. These structural and legal constraints meaningfully protect the human role.
Adoption barriers	claude-sonnet-5	3/5	No strict licensing requirement blocks AI use, but rigorous scientific validation, peer review, and lab safety protocols create substantial institutional friction against full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI tools for analysis and modeling reduce costs on specific subtasks (computational modeling, sequence search), but a biochemist's integrated study activity—experimental design, wet-lab work, interpretation, hypothesis refinement—remains labor-intensive and cheaper to execute with skilled humans than to automate end-to-end with current systems plus oversight infrastructure.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the wet-lab and experimental work central to this task, so there is no viable cost comparison for full automation; human researchers remain necessary.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product autonomously performs biochemical research end-to-end. AI tools exist for protein structure prediction (AlphaFold), sequence analysis, and literature mining, but these are components of the broader study task, not standalone solutions that replace the biochemist's investigative work. Production use remains limited to specific subtasks under expert supervision.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously conducts biochemical research on living processes; AI is used only for narrow sub-components like literature review or data analysis, not the full task.

Develop or test new drugs or medications intended for commercial distribution.

CI 16–28 · exposure 25 · augmentation 75 · importance 3.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Pharma and biotech are adopting AI for early-stage screening and target identification, but clinical development and regulatory submission remain human-led. Adoption of AI tools is present but slow in replacing core drug development roles; most deployment is assistive rather than substitutive.
Sector adoption velocity	claude-sonnet-5	3/5	Pharma and biotech are adopting AI tools for target identification and molecule design at a growing pace, but broader drug development workflows still show pilot-stage rather than deep production-level integration.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI augments biochemists substantially through molecular modeling, hit-to-lead optimization, biomarker prediction, and data analysis, significantly accelerating hypothesis generation and screening. These tools meaningfully raise productivity while the biochemist remains essential for experimental design, interpretation, and regulatory strategy.
Augmentation potential	claude-sonnet-5	4/5	AI substantially augments biochemists by accelerating hypothesis generation, molecular modeling, literature review, and data analysis, meaningfully boosting productivity while humans retain control over experimental validation and regulatory decisions.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with molecular docking, structure prediction, and initial screening, developing and testing drugs for commercial distribution requires extensive experimental validation, clinical trials, regulatory navigation, and human judgment that cannot be fully automated. Current AI cannot end-to-end replace this work with 50% time savings at equal quality.
Task automatability	claude-sonnet-5	2/5	AI can accelerate specific sub-steps like molecule generation, virtual screening, and literature synthesis, but the overall task includes wet-lab experimentation, clinical trials, and regulatory validation that cannot be automated end-to-end today.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Drug development and commercial distribution are heavily regulated by agencies (FDA, EMA, etc.). Licensed professionals must design experiments, interpret results, and sign off on safety and efficacy; liability for adverse outcomes is substantial and non-transferable to software. Regulatory approval mandates human accountability.
Adoption barriers	claude-sonnet-5	4/5	Drug development for commercial distribution is heavily regulated (FDA/EMA approval, GLP/GMP compliance, clinical trial oversight) requiring licensed scientists and regulatory sign-off, creating strong legal and liability barriers to full automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	AI-assisted drug development still requires expensive laboratory infrastructure, animal/clinical testing, and specialized biochemist oversight. The total cost of developing a drug dwarfs AI inference costs, making AI a minor cost factor rather than an economical replacement for the biochemist's role.
Cost vs. human wage	claude-sonnet-5	2/5	AI can cut costs in early-stage in silico screening substantially, but the dominant costs of drug development—synthesis, preclinical/clinical trials, regulatory compliance—remain human- and lab-intensive, keeping overall cost savings modest.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI tools exist for drug discovery components (e.g., AlphaFold for structure, ML for screening), but no deployed product autonomously develops or tests drugs ready for commercial distribution. Products handle narrow segments; the full pipeline remains human-driven with material experimental and regulatory requirements.
Technical feasibility today	claude-sonnet-5	2/5	Deployed AI tools (e.g., AlphaFold-derived platforms, generative chemistry products) assist in drug discovery pipelines at some biotech firms, but they are narrow components within a much larger human-driven R&D and clinical process, not reliable end-to-end production systems.

Develop new methods to study the mechanisms of biological processes.

CI 7–30 · exposure 13 · augmentation 75 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Academic and biotech research sectors adopt AI slowly for method development; most use is exploratory or pilot-stage. The pace of displacement in core method innovation remains modest because experimentation is intrinsically iterative and requires human scientist involvement, and career incentives still reward personal research contributions.
Sector adoption velocity	claude-sonnet-5	2/5	Life sciences R&D is adopting AI tools (e.g., for literature synthesis, hypothesis generation) but actual production-level use for inventing new experimental methodologies remains rare and pilot-stage.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI significantly augments biochemists in this task by accelerating literature synthesis, suggesting experimental designs, running simulations, and analyzing high-dimensional data—keeping the researcher in the loop while boosting ideation and hypothesis testing velocity. This is a strong augmentation use case even though automation remains limited.
Augmentation potential	claude-sonnet-5	4/5	AI tools can meaningfully assist by synthesizing literature, suggesting experimental designs, modeling molecular mechanisms, and identifying gaps, significantly speeding up the ideation phase even though humans must validate and execute methods.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with literature review, data analysis, and computational modeling of biological processes, conceiving and validating genuinely novel experimental methods requires deep domain expertise, iterative hypothesis refinement, and physical experimentation that current AI cannot fully automate end-to-end. Much of the creative design work and validation still depends on human judgment.
Task automatability	claude-sonnet-5	1/5	Developing novel scientific methods requires original hypothesis generation, deep domain intuition, and iterative physical experimentation that current AI cannot perform end-to-end; AI can assist ideation but not autonomously invent and validate new methodologies.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Institutional review boards, scientific rigor standards, peer review requirements, and the need for human researchers to validate and take responsibility for novel methodologies create significant barriers. New methods must be published and vetted by the scientific community, and regulatory/ethical oversight requires human-expert sign-off.
Adoption barriers	claude-sonnet-5	3/5	No licensing requirement specifically blocks AI from proposing methods, but peer review, lab validation, and scientific credibility norms create substantial friction against trusting AI-generated methodologies without extensive human vetting.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The cost of developing new biological methods—including AI tools, computational infrastructure, and human expert time—remains substantial. A skilled biochemist's salary and lab overhead easily exceed the inference and tooling costs of current AI systems, making this an expensive automation relative to human labor in this domain.
Cost vs. human wage	claude-sonnet-5	1/5	Since AI cannot independently perform this task, there is no viable cost comparison; any AI contribution requires substantial expert oversight, making all-in cost higher than a human doing it alone reliably.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product today reliably develops new experimental methods autonomously. AI tools exist for literature mining, molecular simulation, and experimental design suggestion, but these are assistive rather than capable of producing validated novel methods at production scale without substantial human oversight and iteration.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously develops new biological research methods; this remains a research-stage aspiration even for advanced AI scientific discovery tools like those from DeepMind or AI co-scientist prototypes.

Design or perform experiments with equipment, such as lasers, accelerators, or mass spectrometers.

CI 5–25 · exposure 13 · augmentation 75 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Adoption in this domain is minimal; biochemistry and biophysics remain conservative, highly regulated fields where equipment operation is protected by expertise requirements, safety standards, and institutional inertia. Physical laboratory work has shown slow digitization and minimal AI agent deployment.
Sector adoption velocity	claude-sonnet-5	2/5	Physical science lab work adopts AI more slowly than digital-only fields; automation is mostly confined to data analysis, not equipment operation, though robotic lab automation is growing in some sectors.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI provides substantial assistance through experimental design optimization, parameter prediction, automated data analysis, and literature-informed protocol suggestions—all of which can accelerate and improve a biochemist's experimental workflow while they retain full control over equipment and decision-making.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids experimental design, data interpretation, anomaly detection, and literature synthesis, improving productivity even though physical operation remains human-led.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist in experimental design and data analysis, the hands-on operation of specialized equipment (lasers, accelerators, mass spectrometers) and real-time troubleshooting remain fundamentally human tasks requiring physical presence and adaptive decision-making. Current AI systems lack the embodied capability and contextual judgment to perform end-to-end experimental work at the required quality and safety standards.
Task automatability	claude-sonnet-5	1/5	Physical experimentation with lab equipment like lasers, accelerators, or mass spectrometers requires manual manipulation, calibration, and real-time troubleshooting that AI cannot perform end-to-end today.'
Adoption barriers	claude-haiku-4-5-20251001	4/5	Significant regulatory and liability barriers exist: equipment operation often requires certification, institutional safety protocols demand human accountability, and experimental validation has legal/publication consequences. Insurance and institutional review boards typically mandate human responsibility for equipment use and data integrity.
Adoption barriers	claude-sonnet-5	4/5	Operating hazardous or costly scientific equipment typically requires trained, credentialed personnel, safety certifications, and institutional oversight, creating strong organizational and regulatory barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	The cost of integrating AI for equipment operation, including safety systems, oversight, and the requirement for human supervision due to equipment complexity and hazard potential, remains substantially higher than the marginal cost of having a trained biochemist perform the task themselves.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical labor and specialized instrument operation involved, so there is no meaningful cost comparison—human scientists and technicians remain necessary.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI systems can generate experimental protocols and optimize parameters in silico, but no deployed product reliably executes complete wet-lab or instrument-intensive experiments independently. Deployed systems exist only for narrow subtasks (protocol suggestion, spectrum interpretation) rather than the full experiment lifecycle including equipment operation and real-time adjustment.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product autonomously operates or designs full physical experiments with such equipment; this remains research-stage (e.g., 'AI scientist' prototypes) rather than production reality.

Study physical principles of living cells or organisms and their electrical or mechanical energy, applying methods and knowledge of mathematics, physics, chemistry, or biology.

CI 7–21 · exposure 5 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	While computational biology and AI-assisted analysis are growing in research settings, adoption remains pilot-heavy and limited to supporting tools rather than autonomous task performance. Deep scientific work still moves slowly in adoption of fully autonomous systems.
Sector adoption velocity	claude-sonnet-5	2/5	Academic and biomedical research settings adopt AI tools moderately for literature review, data analysis, and modeling, but experimental and conceptual research work still proceeds largely unautomated.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI assists meaningfully on specific components—literature search, data visualization, statistical analysis of experimental results—but does not transform the core creative work of formulating testable hypotheses about cellular physics or interpreting results.
Augmentation potential	claude-sonnet-5	4/5	AI substantially aids literature synthesis, statistical modeling, simulation, and hypothesis generation, meaningfully boosting researcher productivity even though humans remain central to the scientific process.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires novel experimental design, hypothesis formation, and interpretation of complex biological phenomena—core scientific reasoning that current AI cannot perform end-to-end. While AI can assist in data analysis or literature review, the fundamental creative and interpretive work of studying physical principles remains beyond autonomous AI capability.
Task automatability	claude-sonnet-5	1/5	This describes open-ended scientific investigation requiring hypothesis generation, experimental design, wet-lab work, and novel physical/biological reasoning that current AI cannot perform end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	This task requires human scientific judgment, peer review accountability, and institutional responsibility for novel findings. Regulatory and organizational norms expect human scientists to own hypothesis formation and interpretation; liability and credibility barriers are substantial.
Adoption barriers	claude-sonnet-5	3/5	While no formal licensing is required to do research, institutional review, funding accountability, and scientific credibility standards create friction against full automation of research authorship and design.
Cost vs. human wage	claude-haiku-4-5-20251001	2/5	AI can reduce costs on specific subtasks (data processing, literature mining) but cannot replace the full scope of biochemist labor, which commands high expertise-dependent wages. AI support is supplementary, not cost-displacing at the task level.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the human researcher's core scientific work, so there is no meaningful cost-equivalent comparison for the full task.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	No deployed product performs this integrative task end-to-end. AI tools can support narrow components (molecular dynamics simulation, image analysis of cells) but no production system independently studies physical principles of organisms or designs new experimental approaches to understand cellular energy.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously conducts biophysical research on living systems; AI tools are limited to narrow analytical support like data processing or literature review.

Produce pharmaceutically or industrially useful proteins, using recombinant DNA technology.

CI 7–21 · exposure 13 · augmentation 75 · importance 3.2/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	2/5	Biotech and pharma sectors show modest adoption of AI for computational design and process modeling, but automation of actual protein synthesis remains limited to narrow, pre-defined bioreactor runs. Most production still depends on skilled human technicians and scientists making real-time adjustments.
Sector adoption velocity	claude-sonnet-5	2/5	Biotech/pharma manufacturing is a physical, highly regulated sector with slower AI adoption for hands-on production tasks, though AI is increasingly used in upstream design and simulation.
Augmentation potential	claude-haiku-4-5-20251001	4/5	AI strongly augments this task by automating protein sequence design, predicting optimal expression conditions, and analyzing structural data, allowing biochemists to focus on experimental validation and process optimization. These tools meaningfully increase productivity while keeping humans as essential decision-makers.
Augmentation potential	claude-sonnet-5	4/5	AI tools significantly assist with protein structure prediction, codon optimization, vector design, and experimental planning, meaningfully boosting researcher productivity even though the physical production remains human-performed.
Task automatability	claude-haiku-4-5-20251001	2/5	While AI can assist with protein sequence design and some computational steps, the physical production of recombinant proteins requires hands-on laboratory work, bioreactor management, and troubleshooting that cannot be fully automated end-to-end with current systems. Significant human oversight and intervention remain necessary for cloning, expression optimization, and quality control.
Task automatability	claude-sonnet-5	1/5	This is a hands-on wet-lab task involving cloning, transformation, fermentation, purification, and quality control that requires physical manipulation of biological materials; AI cannot perform the physical steps end-to-end.aracts.: current systems can only assist with design/analysis portions.rating rationale limited to 1. (remove filler)
Adoption barriers	claude-haiku-4-5-20251001	4/5	Pharmaceutical protein production is heavily regulated (FDA, EMA oversight), requires quality assurance sign-off by trained scientists, involves biosafety protocols, and often mandates human accountability for batch reproducibility and safety. Regulatory requirements effectively mandate human involvement and certification.
Adoption barriers	claude-sonnet-5	4/5	Producing pharmaceutical proteins is heavily regulated (GMP, FDA oversight, biosafety), requiring qualified scientists and validated physical processes, creating strong institutional and regulatory barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The infrastructure and labor costs for recombinant protein production (equipment, materials, skilled technicians, facilities) far exceed current AI inference costs, and automation of the entire pipeline would require significant custom integration investment that keeps total cost well above human-performed work.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the lab equipment, reagents, and physical labor needed for protein expression and purification, so it cannot be cheaper than human-run wet-lab work since it can't perform the task at all.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	AI tools exist for computational protein design and sequence optimization, but no deployed product reliably performs the complete wet-lab production of recombinant proteins independently. Current systems lack the embodied manipulation, real-time process monitoring, and adaptive problem-solving needed for this multi-step synthesis task at production scale.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs the actual wet-lab production of recombinant proteins; AI tools exist only for design/sequence optimization stages, not the physical production process.

Manage laboratory teams or monitor the quality of a team's work.

CI 5–16 · exposure 8 · augmentation 38 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Adoption of AI for team management in research settings remains minimal. Biochemistry and biophysics labs rely on senior scientists and PI judgment for quality control and personnel oversight, with minimal movement toward automation of these leadership functions.
Sector adoption velocity	claude-sonnet-5	2/5	While AI adoption in life sciences R&D is growing for data analysis, adoption of AI for actual team management functions remains minimal and slow.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI can assist with data aggregation and metric visualization for lab performance, but offers limited augmentation for the core interpersonal and evaluative aspects of team management. The augmentation potential is narrow, focused on administrative support rather than enhancing managerial judgment.
Augmentation potential	claude-sonnet-5	3/5	AI can help by tracking experiment metadata, flagging anomalies in data quality, summarizing lab notebooks, and drafting performance reports, aiding but not replacing the managerial oversight role.
Task automatability	claude-haiku-4-5-20251001	2/5	AI lacks the contextual judgment and interpersonal authority needed to manage teams or conduct meaningful performance evaluations. While AI could generate status reports or flag quality metrics, it cannot replace the human decision-making required for personnel management, feedback, and team coordination.
Task automatability	claude-sonnet-5	1/5	Managing people, setting priorities, mentoring, and resolving interpersonal/team issues require human judgment, authority, and relationship management that current AI cannot perform end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	4/5	Strong organizational and professional barriers exist: management roles require institutional authority, direct accountability for personnel decisions, and organizational trust. Lab directors and PIs have legal responsibility for team performance and scientific integrity that cannot be delegated to automated systems.
Adoption barriers	claude-sonnet-5	4/5	Organizational accountability, HR/legal responsibility for personnel decisions, and the need for a qualified human supervisor to sign off on lab quality/compliance create strong barriers to automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	Implementing an AI system to handle team management would require substantial oversight and validation by human managers, negating any cost advantage. The loaded cost of human management infrastructure plus AI integration would exceed the cost of human managers performing these duties directly.
Cost vs. human wage	claude-sonnet-5	1/5	There is no AI substitute performing the managerial role itself, so cost comparison favors the human who must be employed regardless.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI product today reliably performs laboratory team management or quality oversight as a standalone system. Existing tools may assist with data tracking or reporting, but the core task of evaluating scientist performance and managing team dynamics remains fundamentally dependent on human leadership.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product autonomously manages laboratory teams or holistically monitors work quality across a team; this remains a human supervisory function.

Prepare pharmaceutical compounds for commercial distribution.

CI 9–11 · exposure 16 · augmentation 50 · importance 3.8/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Pharmaceutical manufacturing and quality control remain heavily human-dependent sectors with strict regulatory oversight that actively discourages full automation without human sign-off. Adoption of AI in this domain is limited to supporting roles (data analysis, documentation) rather than autonomous task execution.
Sector adoption velocity	claude-sonnet-5	2/5	Pharma manufacturing is a physically intensive, highly regulated sector where AI adoption for core production tasks lags well behind adoption in data-centric fields, though AI is used in some analytics and quality support roles.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can meaningfully assist biochemists by automating data analysis, predicting formulation stability, managing batch documentation, and flagging out-of-specification results, raising their efficiency in reviewing and decision-making aspects of compound preparation.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with process optimization, predictive quality control, documentation, and batch record analysis, improving efficiency without replacing physical compounding operations.
Task automatability	claude-haiku-4-5-20251001	2/5	Preparing pharmaceutical compounds for commercial distribution requires precise formulation, quality control testing, stability analysis, and regulatory compliance—tasks that demand human judgment, complex decision-making, and accountability. While AI could assist with documentation, batch scheduling, or data analysis, it cannot autonomously handle the full end-to-end process at the required quality and safety standards.
Task automatability	claude-sonnet-5	2/5	This task involves physical formulation, GMP manufacturing processes, quality control testing, and regulatory compliance activities that require hands-on lab and plant work AI cannot perform.; only documentation and some data analysis portions are automatable.
Adoption barriers	claude-haiku-4-5-20251001	5/5	Pharmaceutical manufacturing is heavily regulated by FDA, EMA, and other agencies requiring licensed chemists or engineers to certify compound preparation, perform release testing, and sign quality assurance documentation. Legal liability, cGMP compliance, and the necessity of human expert judgment and accountability create hard regulatory barriers to full automation.
Adoption barriers	claude-sonnet-5	5/5	Pharmaceutical manufacturing is heavily regulated by agencies like the FDA under GMP requirements, requiring qualified personnel, validated processes, and documented sign-offs, making unsupervised AI substitution legally impossible.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The cost of developing and operating AI systems capable of performing pharmaceutical manufacturing and quality assurance, combined with mandatory human oversight and liability, would exceed the loaded wage of experienced biochemists and biopharmaceuticists who currently perform this work.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical compounding, mixing, and quality-controlled manufacturing equipment and labor required, so there is no meaningful cost comparison for the core task.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed product reliably performs pharmaceutical compound preparation for commercial distribution independently. This remains a highly regulated, hands-on process requiring trained chemists and biopharmaceutical engineers to handle synthesis, sterility assurance, potency verification, and regulatory sign-off in production settings.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product manufactures or physically prepares pharmaceutical compounds; this remains a physical, equipment- and facility-dependent process governed by strict manufacturing protocols.

Teach or advise undergraduate or graduate students or supervise their research.

CI 0–13 · exposure 5 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Academic institutions are laggards in automation; teaching and mentoring remain highly protected by institutional mission and governance. Adoption of AI for core teaching roles is negligible in production settings.
Sector adoption velocity	claude-sonnet-5	2/5	Higher education adopts AI tools for research assistance and tutoring support at a moderate pace, but the core advising/supervision function itself sees minimal AI-driven displacement.
Augmentation potential	claude-haiku-4-5-20251001	3/5	AI can assist with grading, generating practice problems, providing preliminary literature summaries, and answering routine questions, thereby freeing faculty time for deeper mentorship. However, the augmentation is incremental rather than transformative given that faculty remain the essential bridge.
Augmentation potential	claude-sonnet-5	4/5	AI can meaningfully assist with literature searches, drafting feedback on writing, generating practice problems, and explaining concepts, enhancing an advisor's efficiency while they remain central to mentorship.
Task automatability	claude-haiku-4-5-20251001	1/5	Teaching and advising students requires real-time interaction, relationship-building, and adaptive responses to individual learning needs. Current AI systems cannot replicate the mentoring relationship or the nuanced judgment needed to guide research direction and supervise student progress at equal quality.
Task automatability	claude-sonnet-5	1/5	Mentoring, advising, and supervising original research requires relational judgment, personalized feedback, and accountability that current AI cannot replicate end-to-end.」
Adoption barriers	claude-haiku-4-5-20251001	5/5	Universities have strong institutional, accreditation, and regulatory requirements that faculty directly teach and mentor students. The human-contact requirement and fiduciary duty to students create hard barriers to substitution of this core educational function.
Adoption barriers	claude-sonnet-5	4/5	Academic institutions require credentialed faculty to formally advise and supervise research, sign off on theses, and be accountable for training outcomes, creating strong institutional and credentialing barriers.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	The cost of integrating AI tutoring systems with appropriate oversight, plus the need for human faculty to remain engaged in actual supervision and mentorship, means total cost remains higher than current university labor economics.
Cost vs. human wage	claude-sonnet-5	1/5	There is no viable AI-only substitute for this task, so cost comparison is moot; any attempt would require extensive human oversight negating cost savings.
Technical feasibility today	claude-haiku-4-5-20251001	2/5	While AI can generate explanatory content or answer factual questions, no deployed system reliably supervises student research, provides personalized mentorship, or assesses conceptual understanding with the depth required. Chatbots can assist but cannot replace the core teaching and advising function.
Technical feasibility today	claude-sonnet-5	1/5	No deployed product performs student advising or research supervision autonomously; AI tools exist only as supplementary aids for tutoring or literature review, not as substitutes for the advisory relationship.

Research transformations of substances in cells, using atomic isotopes.

CI 0–7 · exposure 0 · augmentation 38 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
Dimension	Model	Rating	Rationale
Sector adoption velocity	claude-haiku-4-5-20251001	1/5	Biochemistry and isotope research remain traditional, small-team, regulation-heavy fields with low digitization and minimal AI adoption patterns; institutional and safety barriers slow any technological transition.
Sector adoption velocity	claude-sonnet-5	2/5	Academic and biomedical research labs adopt AI tools for data analysis and literature review, but the physical experimental core of isotope tracer work sees minimal AI-driven displacement.
Augmentation potential	claude-haiku-4-5-20251001	2/5	AI could modestly assist with literature review, data analysis of isotope-tracing results, or computational modeling of reaction pathways, but the core experimental design, isotope handling, and real-time laboratory work remain fundamentally human-driven.
Augmentation potential	claude-sonnet-5	3/5	AI can assist with experimental design, data analysis of isotope-tracing results, and interpreting metabolic flux data, meaningfully aiding the surrounding analytical work even though the physical assay itself is unaffected.
Task automatability	claude-haiku-4-5-20251001	1/5	This task requires hands-on laboratory work with radioactive isotopes, specialized equipment operation, and real-time adaptive decision-making based on experimental observations. Current AI systems cannot physically perform isotope work or reliably design and execute novel biochemical experiments without human intervention.
Task automatability	claude-sonnet-5	1/5	This is hands-on wet-lab experimental work involving radiolabeling, physical sample handling, and instrumentation that AI cannot physically perform; only ancillary data analysis could be offloaded.the core task remains unautomatable end-to-end.
Adoption barriers	claude-haiku-4-5-20251001	5/5	This task is heavily regulated by licensing requirements (radiation safety, institutional review boards, specialized credentials), legal liability for isotope handling errors, and inherent safety/authorization rules that mandate human professional responsibility and sign-off.
Adoption barriers	claude-sonnet-5	4/5	Radioisotope handling requires specialized licensing, safety training, and regulatory compliance (e.g., radiation safety officers), plus the physical nature of the work creates strong barriers to any automation.
Cost vs. human wage	claude-haiku-4-5-20251001	1/5	Deploying AI to this task would require expensive custom hardware integration, specialized lab robotics, and extensive human oversight; the cost would exceed that of a trained biochemist performing the work directly.
Cost vs. human wage	claude-sonnet-5	1/5	AI cannot substitute for the physical experimental process, so there is no meaningful AI cost basis to compare against the human-performed wet-lab work.
Technical feasibility today	claude-haiku-4-5-20251001	1/5	No deployed AI system can independently research cellular transformations using atomic isotopes; this remains fundamentally a research and experimental domain requiring human scientists to design protocols, operate instruments, and interpret complex biological data.
Technical feasibility today	claude-sonnet-5	1/5	No deployed AI product conducts isotope-tracer experiments in cells; this remains firmly in the domain of physical laboratory science requiring human execution and specialized equipment.

Related occupations — Life, Physical & Social Science

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.