Healthcare Social Workers
21-1022.00Median wage $67,880/yr187,630 employed (US)Rank #733 of 923 scored · top 79% by substitution
Provide individuals, families, and groups with the psychosocial support needed to cope with chronic, acute, or terminal illnesses. Services include advising family caregivers. Provide patients with information and counseling, and make referrals for other services. May also provide case and care management or interventions designed to promote health, prevent disease, and address barriers to access to healthcare.
Sub-scores
0–100 · band = confidence interval from rater disagreement
Substitution18
Exposure16
Augmentation56
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
17 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
0%
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%17
panel mean rating 1.7/5 → substitution pressure 17/100
Technical feasibility todayw 20%15
panel mean rating 1.6/5 → substitution pressure 15/100
Cost vs. human wagew 15%21
panel mean rating 1.9/5 → substitution pressure 21/100
Adoption barriersw 20%inverted — strong barriers lower the score17
panel mean rating 4.3/5 (barrier strength) → substitution pressure 17/100
Sector adoption velocityw 10%20
panel mean rating 1.8/5 → substitution pressure 20/100
Task breakdown (17 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.
Oversee Medicaid- and Medicare-related paperwork and recordkeeping in hospitals.
57CI 45–70 · exposure 62 · augmentation 75 · importance 3.5/5 · click for rater detail
Oversee Medicaid- and Medicare-related paperwork and recordkeeping in hospitals.
57| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 4/5 | Healthcare systems and hospital billing departments are rapidly adopting RPA and AI-based document processing; major EHR vendors (Epic, Cerner) integrate automated compliance checking. Production adoption is measurable and accelerating across hospital networks, though lagging in smaller or rural facilities. |
| Sector adoption velocity | claude-sonnet-5 | 3/5 | Healthcare administration is adopting AI documentation and RCM tools at a moderate pace, with pilots and partial deployments common but full-scale replacement still limited by regulatory caution. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI systems that flag errors, auto-populate forms, validate coding against regulations, and alert to missing documents substantially augment a social worker's ability to review Medicaid/Medicare paperwork more quickly and accurately, keeping the human in the oversight loop. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI substantially aids drafting, coding, and tracking of Medicaid/Medicare records, reducing administrative burden while the social worker retains responsibility for accuracy and compliance decisions. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | A large portion of Medicaid and Medicare paperwork—document submission, eligibility verification, coding checks, and compliance logging—can be automated with existing workflow systems and document-processing AI, easily achieving 50% time savings at equal or better quality. However, complex case adjudication and appeals still require human judgment, preventing a full 5. |
| Task automatability | claude-sonnet-5 | 3/5 | AI can draft, code, and organize much of the paperwork and flag documentation gaps, but final oversight, exception handling, and compliance judgment still require human review, limiting full automation to roughly half the workflow. |
| Adoption barriers | claude-haiku-4-5-20251001 | 3/5 | While Medicaid and Medicare regulations govern the data itself, there is no legal requirement that a licensed social worker (rather than a health information specialist or algorithm) perform the recordkeeping oversight. However, hospitals maintain internal compliance review and audit requirements, and liability concerns over incorrect submissions create modest friction against full replacement. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Medicaid/Medicare compliance carries strict regulatory, audit, and liability requirements, often necessitating credentialed social worker or compliance officer sign-off, creating strong institutional and legal barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | Cloud-based document processing, RPA, and healthcare compliance automation are considerably cheaper than the fully loaded cost of a social worker's time on routine paperwork tasks—likely 3–10× cheaper for high-volume reconciliation and data entry. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted documentation and claims-processing tools reduce labor time but still require licensed oversight and integration costs, making savings meaningful but not order-of-magnitude cheaper than human labor alone. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 4/5 | Deployed products (RPA, healthcare-specific document processing, compliance software, and billing automation platforms) reliably handle Medicaid/Medicare recordkeeping in hospital production environments. Error rates remain non-zero on edge cases, but the mainstream tasks have mature solutions at scale. |
| Technical feasibility today | claude-sonnet-5 | 3/5 | Deployed EHR-integrated tools and RCM/compliance software (e.g., NLP-based coding assistants) exist in production hospitals, but error rates and need for human verification of Medicaid/Medicare rules remain material. |
Monitor, evaluate, and record client progress according to measurable goals described in treatment and care plan.
40CI 25–55 · exposure 45 · augmentation 75 · importance 4.3/5 · click for rater detail
Monitor, evaluate, and record client progress according to measurable goals described in treatment and care plan.
40| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare social work remains in laggard-to-middling adoption of automation; most agencies still rely on manual charting and legacy systems, with only larger healthcare systems piloting AI-assisted documentation tools, reflecting organizational inertia and regulatory caution. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare and social services sectors show slower, more cautious AI adoption due to regulatory, ethical, and workforce factors compared to fast-adopting sectors like finance or tech. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI assistants demonstrably boost productivity by auto-populating progress notes, flagging missed metrics, and cross-referencing outcomes against plan goals, allowing clinicians to focus on synthesis and client relationship while the system handles routine data alignment and recording. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist with note-taking, summarizing session data, flagging goal deviations, and drafting documentation, improving efficiency while the social worker retains judgment and oversight. |
| Task automatability | claude-haiku-4-5-20251001 | 4/5 | AI can systematically extract clinical data from health records, compare documented outcomes against established metrics, and generate progress reports with high consistency, though human judgment on goal relevance and plan adjustments typically remains necessary. This meets the 50% time-saving threshold for the monitoring and recording components. |
| Task automatability | claude-sonnet-5 | 2/5 | Recording and summarizing progress notes against goals can be partially automated with AI transcription/drafting tools, but ongoing clinical evaluation and judgment about client progress require human assessment and cannot be fully offloaded today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong regulatory and liability barriers exist: clinicians are legally responsible for treatment plan accuracy and progress documentation in most jurisdictions; HIPAA compliance and clinical oversight requirements limit autonomous automation; and client confidentiality mandates human sign-off on records. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Clinical documentation and progress evaluation typically require licensed professional sign-off, with regulatory, confidentiality (HIPAA), and liability constraints limiting full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 4/5 | AI-driven charting and outcome tracking systems have substantially lower marginal costs per client than human administrative time, though setup and oversight costs mean the ratio is favorable but not yet an order of magnitude cheaper. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI documentation assistance can reduce some administrative time, but human clinical oversight, verification, and judgment remain necessary, keeping all-in costs roughly comparable to human-only work with modest savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 3/5 | Products exist for electronic health record integration, outcome tracking, and automated progress note generation, but real-world deployment shows inconsistencies in handling complex social factors and requires clinician review before finalization, limiting full end-to-end deployment reliability. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI scribe and documentation tools exist and are used in some healthcare settings, but reliable, validated systems for evaluating psychosocial progress against care plans in social work contexts are not widely deployed. |
Refer patient, client, or family to community resources to assist in recovery from mental or physical illness and to provide access to services such as financial assistance, legal aid, housing, job placement or education.
29CI 25–34 · exposure 25 · augmentation 63 · importance 4.5/5 · click for rater detail
Refer patient, client, or family to community resources to assist in recovery from mental or physical illness and to provide access to services such as financial assistance, legal aid, housing, job placement or education.
29| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare and social services sectors are historically slow in automation adoption, with heavy reliance on licensed practitioners, regulatory constraints, and resistance to depersonalization of care. While digital tools are being piloted, production-level autonomous referral systems remain rare. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare and social services adopt AI more slowly due to regulatory, ethical, and funding constraints, though care coordination software adoption is growing modestly. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by rapidly identifying candidate resources, flagging eligibility criteria, and compiling information summaries that social workers review and refine. However, the core judgment and relationship work remains with the human. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can significantly speed up resource identification, eligibility matching, and documentation, letting social workers focus more on relationship-based aspects of referral and follow-up. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can gather and organize information about available community resources and match them to patient needs, the task requires nuanced judgment about individual circumstances, trustworthiness of resources, and cultural fit that current systems struggle with reliably. The referral decision involves significant human discretion and relationship factors that cannot be automated end-to-end to the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can help identify and compile relevant resource lists, but the actual referral requires understanding nuanced patient circumstances, building trust, and coordinating with agencies, which limits full end-to-end automation. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist due to licensure requirements for social workers in most jurisdictions, liability and malpractice exposure if referrals prove inappropriate or harmful, duty-of-care obligations, and legal requirements that a qualified professional assess and vouch for fit. Patient vulnerability and regulatory oversight of social services create substantial friction. |
| Adoption barriers | claude-sonnet-5 | 3/5 | No strict licensing requirement for referral itself, but liability, confidentiality (HIPAA), and the need for professional judgment in vulnerable populations create moderate friction against full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI systems for resource matching require significant integration, curation, and oversight to be useful. The cost of maintaining accurate databases, handling exceptions, and human verification approaches or exceeds the cost of a social worker performing this task directly. |
| Cost vs. human wage | claude-sonnet-5 | 3/5 | AI-assisted resource lookup is cheap compared to a social worker's time, but human judgment, relationship-building, and follow-up still require paid staff, keeping overall cost comparable rather than order-of-magnitude lower. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs the full referral task independently. Existing systems offer resource databases and matching tools, but they operate with material limitations in coverage, accuracy, and contextual appropriateness, and still require substantial human oversight and verification. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some resource-navigation chatbots and databases exist (e.g., 211 systems, care coordination software), but they are narrow tools that assist rather than fully perform personalized referrals reliably. |
Conduct social research to advance knowledge in the social work field.
28CI 25–30 · exposure 25 · augmentation 75 · importance 3.0/5 · click for rater detail
Conduct social research to advance knowledge in the social work field.
28| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Research institutions adopt AI *tools* (statistics, literature management) incrementally and cautiously, but original knowledge advancement remains anchored to human scholars. Adoption of AI-driven research practices is slower in social work than in tech or finance, reflecting conservative academic and ethical norms. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Social work and academic research settings show slower, more cautious AI adoption compared to fast-moving tech or finance sectors, with pilots for literature review tools more common than full research automation. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI substantially augments research productivity: literature review and meta-analysis, data analysis, manuscript drafting, and visualization are all meaningfully accelerated. A human researcher leveraging AI assistants can conduct research faster and explore more hypotheses while maintaining full intellectual control and accountability. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI significantly aids tasks like literature reviews, survey design suggestions, qualitative data coding, and statistical analysis, meaningfully boosting researcher productivity while humans retain control over methodology and conclusions. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Social research requires design of novel studies, ethical judgment in methodology, and interpretation grounded in complex human contexts. While AI can assist with literature review, data analysis, and synthesis, the core task of *advancing knowledge* through original research design and expert interpretation remains fundamentally dependent on human researcher judgment and creativity. |
| Task automatability | claude-sonnet-5 | 2/5 | AI can assist with literature review, data analysis, and drafting portions of social research, but designing studies, ensuring ethical human-subjects protections, and interpreting findings within social work practice require human judgment that current AI cannot fully replace end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Institutional review boards (IRBs), ethical oversight requirements, and professional licensing/accountability in research create regulatory barriers. Published research in peer-reviewed journals requires human author accountability and institutional affiliation, making full automation legally and ethically difficult. |
| Adoption barriers | claude-sonnet-5 | 3/5 | IRB/ethics approval, professional standards for research integrity, and accountability for published findings create moderate barriers, though no strict licensing requirement mandates a human perform every research step. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI tools for research support (literature indexing, analysis) are increasingly affordable, but the human researcher cost remains dominant because expertise, IRB approvals, and research validation cannot be replaced by inference alone. Overall cost savings are modest relative to the human's total effort. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI tools reduce time on literature searches and data processing, but the overall research process still requires substantial human labor for design, fieldwork, and interpretation, keeping costs comparable to human-led research with AI assistance. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed AI product conducts original social research end-to-end; research is still in conceptual stages. AI can support components (data cleaning, statistical summaries) but conducting the full research cycle—from hypothesis to publication—still requires human researchers to own the intellectual direction and accountability. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Research-assistant tools (literature synthesis, statistical analysis software with AI features) are deployed but no product independently conducts rigorous social work research reliably without heavy human oversight and validation. |
Utilize consultation data and social work experience to plan and coordinate client or patient care and rehabilitation, following through to ensure service efficacy.
25CI 25–25 · exposure 25 · augmentation 75 · importance 4.3/5 · click for rater detail
Utilize consultation data and social work experience to plan and coordinate client or patient care and rehabilitation, following through to ensure service efficacy.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare and social services remain digitally fragmented and conservative in automation of care design. Adoption is pilot-heavy; most real production use remains data aggregation and scheduling support, not end-to-end planning automation. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare and social services adopt AI slowly due to regulatory, ethical, and interpersonal trust concerns, with pilots more common than full production use. |
| Augmentation potential | claude-haiku-4-5-20251001 | 4/5 | AI strongly augments this task by rapidly synthesizing consultation notes, flagging service gaps, organizing referral options, and tracking client progress—freeing the social worker to focus on assessment, relationship-building, and real-time care adjustment. Productivity gains are substantial while the human remains accountable. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can meaningfully assist by summarizing consultation data, flagging risks, and drafting care plans, improving efficiency while the social worker retains responsibility for judgment and follow-up. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can help aggregate consultation data and suggest service options, the core task requires balancing client context, dignity, and individualized judgment to design care plans. The follow-through and real-time adjustment based on client feedback and social dynamics remain fundamentally human-dependent; no AI system today automates the full planning-to-verification cycle at 50% time savings with equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Care planning requires synthesizing clinical, psychosocial, and family context with professional judgment and follow-through accountability, which current AI cannot autonomously execute end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: social work care planning is often legally scoped to licensed or regulated professionals; liability for poor outcomes rests with the social worker; and client relationship trust is a documented requirement for efficacy. Many jurisdictions require social worker sign-off on care plans. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Healthcare social work often involves licensure, confidentiality (HIPAA), liability for patient outcomes, and required human judgment in care decisions, creating strong regulatory and professional barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Current AI tools for health coordination (chatbots, data aggregation) incur significant integration and human oversight costs and do not yet replace the clinical judgment labor. The cost per meaningful care plan remains above the loaded wage of a junior social worker. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply draft notes or summarize data, but the full coordination and follow-through role still requires a paid professional, keeping overall cost savings modest. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed system reliably performs end-to-end care planning and coordination for social work clients at scale. AI can assist with data synthesis and referral matching, but actual care plan design and efficacy follow-up require licensed social workers to assess complex psychosocial factors and make accountability decisions. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some care coordination software and AI-assisted documentation tools exist, but no deployed product independently plans and monitors patient care with reliable efficacy tracking. |
Identify environmental impediments to client or patient progress through interviews and review of patient records.
25CI 25–25 · exposure 25 · augmentation 63 · importance 4.2/5 · click for rater detail
Identify environmental impediments to client or patient progress through interviews and review of patient records.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare social work remains a relationship-intensive, documentation-heavy field with slower digitization than billing or scheduling. While EHR adoption is widespread, AI-driven clinical decision support in social work assessment is in early pilot phases; most agencies rely on clinician-led interviews and manual record review. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare and social services generally lag in AI adoption for direct clinical/psychosocial assessment tasks, with most current use limited to administrative documentation support. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by extracting and organizing information from records, flagging previously mentioned barriers, and generating summary prompts to guide interviews—useful productivity aids that help clinicians organize thought. However, the interpretive and relational core of identifying *which* impediments truly matter to this client remains the clinician's work. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can assist by summarizing patient records, highlighting relevant history, and drafting notes, meaningfully speeding up the preparatory work around interviews even though the interview and judgment remain human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | Identifying impediments requires nuanced interpretation of client narratives and contextual factors across interviews and records. While AI can extract and categorize explicit barriers from text, distinguishing true environmental impediments from symptoms, assessing their relative importance, and contextualizing them within a client's unique situation requires sustained clinical judgment that current systems cannot reliably perform end-to-end. |
| Task automatability | claude-sonnet-5 | 2/5 | This requires nuanced synthesis of interview cues, tone, patient history, and social context to identify environmental barriers—judgment-heavy work that current AI cannot reliably perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers protect this task: social workers are licensed professionals whose assessment directly influences treatment planning and client safety; regulatory bodies (NASW, state licensure) expect documented professional judgment; liability for incorrect impediment identification falls on the organization; and clients expect human clinical understanding of sensitive social and environmental contexts. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Healthcare social work involves licensure, confidentiality (HIPAA), and professional judgment requirements that make full delegation to AI legally and ethically constrained. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI-assisted extraction and summarization tools exist but require significant clinical oversight and validation. The total cost (tool license, clinician review time, error correction) remains comparable to or higher than direct clinician assessment, especially given liability concerns in misidentifying impediments. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Because a licensed social worker must still conduct interviews and validate findings, AI only offsets a small documentation/review portion, limiting overall cost savings versus the human's full-service loaded wage. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | NLP systems can retrieve and summarize information from records and flag potential impediments, but no deployed product reliably performs the full diagnostic task—understanding which barriers are primary causes of stalled progress versus secondary factors—with the accuracy required in clinical practice. Existing tools support rather than replace this judgment. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | AI tools can summarize records and flag keywords, but no deployed product independently conducts clinical interviews and derives environmental impediment assessments in production social work settings. |
Plan discharge from care facility to home or other care facility.
25CI 25–25 · exposure 25 · augmentation 63 · importance 4.1/5 · click for rater detail
Plan discharge from care facility to home or other care facility.
25| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare social work remains in relatively early stages of AI adoption, with most systems still using basic digital tools for administrative support rather than AI agents; organizational culture and licensing requirements slow deployment of any replacement-oriented automation. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare is a highly regulated, moderately slow-adopting sector for AI in high-stakes coordination tasks, with pilots more common than production deployment for discharge planning specifically. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by generating resource recommendations, identifying gaps in discharge readiness, and automating paperwork, but the social worker remains essential for assessment, negotiation, and advocacy, making this a genuine augmentation scenario rather than transformation. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI tools can meaningfully assist by drafting discharge summaries, matching patients to facilities, and flagging risks, improving social worker efficiency while the human retains decision-making and coordination responsibility. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist in generating discharge checklists and identifying available resources, discharge planning requires individualized assessment of complex patient needs, family circumstances, and coordination with multiple providers—tasks requiring human judgment and real-time communication that cannot be fully automated to achieve 50% time savings at equal quality. |
| Task automatability | claude-sonnet-5 | 2/5 | Discharge planning requires synthesizing medical, psychosocial, financial, and family circumstances into a judgment-based plan, which current AI cannot execute end-to-end reliably despite being able to assist with parts of it. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Discharge planning is often legally mandated as a social worker responsibility in many jurisdictions, and liability for poor transitions falls on the facility and social worker; patients and families typically expect human-led coordination, creating both regulatory and relational barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Discharge planning often requires licensed social worker or case manager sign-off tied to regulatory and reimbursement requirements (e.g., CMS discharge planning rules), creating significant legal and liability barriers to full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI systems for discharge planning support remain costly to implement and maintain relative to the task's value; the human social worker must still oversee all recommendations, making full substitution unlikely and cost savings marginal compared to the social worker's labor. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI can cheaply generate checklists or summaries, but the coordination, negotiation with families/facilities, and liability oversight still require a paid human professional, keeping overall cost comparable to human labor. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably performs end-to-end discharge planning independently; existing tools offer narrow administrative support (appointment scheduling, resource databases) but lack the ability to assess psychosocial factors, negotiate with families, or coordinate interdisciplinary care at the level required in practice. |
| Technical feasibility today | claude-sonnet-5 | 2/5 | Some hospital systems use AI to flag discharge readiness or predict length of stay, but no deployed product independently plans and coordinates discharge dispositions at scale. |
Plan and conduct programs to combat social problems, prevent substance abuse, or improve community health and counseling services.
16CI 7–25 · exposure 13 · augmentation 50 · importance 3.4/5 · click for rater detail
Plan and conduct programs to combat social problems, prevent substance abuse, or improve community health and counseling services.
16| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare and social services sectors adopt AI slowly for high-stakes tasks; adoption is primarily in back-office functions (scheduling, billing) rather than program planning or service delivery. Substance abuse prevention and community health interventions remain human-led, with AI used only as an assistive tool in pilot settings. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Social services and community health sectors have lower digitization and AI adoption compared to finance or tech, with pilots for administrative support but little for program design and delivery. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by synthesizing research, identifying evidence-based practices, modeling resource allocation, and automating administrative tasks, thereby freeing social workers to focus on community engagement and direct service. However, augmentation is limited to support functions; the core relational and strategic work cannot be substantially accelerated by AI. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with literature review, data analysis on community health trends, drafting program proposals, and organizing resources, meaningfully aiding planning even though execution remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can assist with data analysis, resource identification, and program design drafting, the core task requires establishing trust, conducting needs assessments with community members, and adapting interventions based on human judgment and lived experience—activities that cannot be fully automated. AI might automate 20–30% of administrative prep work but cannot replace the relational and contextual work central to effective program planning and conducting. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires strategic planning, community organizing, stakeholder relationships, and judgment-driven program design that current AI cannot execute end-to-end; AI cannot conduct programs or engage communities autonomously. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Strong barriers exist: regulatory oversight of health and social services, licensure requirements for social workers in many jurisdictions, liability concerns for misguidance on substance abuse or health counseling, and ethical/legal mandates for human accountability in vulnerable populations. Automation would face organizational and legal resistance. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Healthcare social work often requires licensure, ethical accountability, and trust-based human relationships with vulnerable populations, creating strong professional and regulatory barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI can reduce costs for literature searches and initial drafting, but the bulk of the work—community engagement, program delivery, and adaptive management—remains labor-intensive and human-dependent. Cost savings are modest (likely 10–20%), keeping overall cost ratio similar to or slightly better than human labor alone. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | Since AI cannot perform the core planning, community engagement, and program delivery, there is no viable AI substitute cost to compare—human labor remains necessary. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed products reliably perform the full task of planning and conducting community health or substance abuse prevention programs. AI tools exist for literature review, data aggregation, and protocol drafting, but actual program conception, stakeholder engagement, and real-time conduct rely on human expertise and discretion that current systems do not handle reliably at scale. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product plans and runs community health or substance abuse programs; this remains firmly a human relational and administrative task with no production analog. |
Collaborate with other professionals to evaluate patients' medical or physical condition and to assess client needs.
16CI 6–25 · exposure 13 · augmentation 50 · importance 4.6/5 · click for rater detail
Collaborate with other professionals to evaluate patients' medical or physical condition and to assess client needs.
16| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 2/5 | Healthcare remains cautious and highly regulated; while EHR systems and data analytics are widespread, AI agents performing autonomous patient assessment and needs evaluation are rare in production, with adoption driven more by incremental tool adoption than displacement. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare social work remains a relationship-driven, low-digitization field where AI adoption is mostly limited to administrative tools, not the core assessment collaboration. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can meaningfully assist by organizing medical records, highlighting key clinical findings, and generating preliminary assessment summaries, thereby allowing social workers to focus on dialogue and judgment; however, the augmentation is partial and does not transform the core evaluative process. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help by summarizing medical records, flagging risk factors, or drafting assessment notes, giving moderate assistance to the social worker's evaluation process while judgment remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can help summarize medical records and flag relevant clinical data, the core task—collaborative professional evaluation and holistic needs assessment—requires human judgment, interdisciplinary dialogue, and contextual understanding of patient circumstances. Current systems cannot reliably conduct end-to-end evaluations meeting the 50% time-saving threshold. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires interprofessional collaboration, in-person or synchronous case discussion, and integrated clinical-social judgment about a specific patient, which current AI cannot perform end-to-end.atical AI cannot replace this collaborative human process today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Healthcare licensing requirements, liability concerns around clinical assessment errors, and regulatory standards (e.g., HIPAA, professional board oversight) mandate that a qualified social worker conduct and sign off on patient evaluations; AI cannot substitute for this credentialed human function. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Healthcare social work involves licensure, clinical accountability, confidentiality (HIPAA), and multidisciplinary judgment calls that legally and ethically require a qualified human professional's involvement. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI infrastructure and oversight costs are comparable to or exceed the value of the limited automation possible; the human social worker must still conduct the primary evaluation and verification, negating significant cost savings. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | Since AI cannot perform the core collaborative evaluation, any cost comparison is moot except for minor note-taking or summarization support, which offers only marginal cost savings against the full task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | AI tools exist to assist with data aggregation and preliminary assessments (EHR summaries, risk flagging), but no deployed product reliably performs the collaborative professional evaluation itself; systems remain narrow and lack the nuance required for genuine patient needs assessment in production settings. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product autonomously collaborates with clinicians to jointly evaluate a patient's condition and social needs; existing tools are limited to documentation support or triage suggestions. |
Educate clients about end-of-life symptoms and options to assist them in making informed decisions.
12CI 4–20 · exposure 13 · augmentation 50 · importance 4.6/5 · click for rater detail
Educate clients about end-of-life symptoms and options to assist them in making informed decisions.
12| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare and social services sectors are slow to adopt AI in high-touch, high-stakes counseling roles; end-of-life work is especially conservative due to liability and ethical norms, and there is minimal evidence of AI replacing social workers in this function even in digitally mature organizations. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare and social services generally show slower AI adoption for direct client-facing emotional/clinical interactions compared to administrative tasks, with pilots focused on documentation rather than counseling delivery. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI could usefully assist by preparing plain-language fact sheets, summarizing medical options, or flagging resource databases, which a social worker could review and discuss with clients; however, the core task—educating and counseling—remains human-driven, so augmentation is partial rather than transformative. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help social workers prepare educational materials, summarize medical information about prognosis/options, and draft talking points, meaningfully supporting preparation without conducting the sensitive conversation itself. |
| Task automatability | claude-haiku-4-5-20251001 | 2/5 | While AI can generate factual information about end-of-life symptoms and options, the task fundamentally requires personalized dialogue, emotional attunement, and exploration of individual values—domains where current AI systems lack reliability and the stakes are too high for unassisted automation. Meaningful time savings would require replacing the human entirely, which is neither feasible nor acceptable in this context. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires real-time emotional attunement, trust-building, and adapting sensitive information to a distressed client's psychological state, which current AI cannot replicate end-to-end despite being able to generate accurate content on end-of-life topics. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Strong legal, ethical, and regulatory barriers protect this task: licensure requirements for social workers, liability exposure, duty-of-care standards, informed consent obligations, and organizational/patient preference for a licensed human to facilitate such consequential decisions. End-of-life counseling is not automatable by regulation and practice. |
| Adoption barriers | claude-sonnet-5 | 5/5 | This involves high-stakes, emotionally sensitive clinical communication with legal and ethical implications (informed consent, licensure, liability) that mandates a qualified human professional. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | Even if a chatbot deployment were acceptable, the cost of oversight, liability management, and integration into a care workflow would likely exceed the value of the marginal AI assistance, especially given the high error-cost asymmetry in end-of-life decisions. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While generating informational content is cheap, the human relational component requires a licensed social worker present regardless, so AI does not meaningfully reduce the core labor cost of this task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 2/5 | No deployed product reliably handles end-of-life counseling autonomously; AI chatbots exist for health information but are not used in production to replace social workers for this sensitive, legally and ethically fraught conversation. Current systems can provide information but cannot substitute for the judgment and rapport required. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently conducts end-of-life counseling conversations with clients; AI tools exist only as reference or drafting aids for the professional, not as substitutes for the interaction itself. |
Modify treatment plans to comply with changes in clients' status.
10CI 0–20 · exposure 8 · augmentation 63 · importance 4.1/5 · click for rater detail
Modify treatment plans to comply with changes in clients' status.
10| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare remains a heavily regulated, risk-averse sector with strong legal and professional constraints on delegating clinical decisions; adoption of AI for autonomous treatment planning is minimal and unlikely to accelerate soon. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare social work is a moderately digitized but relationship- and compliance-heavy field with slow, cautious AI adoption for clinical decision tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI could assist by summarizing client status changes, flagging relevant data, or drafting plan revisions for the social worker to review and finalize, raising their efficiency without replacing clinical judgment. |
| Augmentation potential | claude-sonnet-5 | 4/5 | AI can help summarize case updates, flag status changes, and draft revised plan language, meaningfully speeding the social worker's review and documentation process. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Modifying treatment plans requires interpreting complex changes in a client's clinical, social, and emotional status, then making nuanced clinical judgments about care. Current AI cannot reliably assess the multifaceted human context needed to legitimately revise a treatment plan. |
| Task automatability | claude-sonnet-5 | 2/5 | Requires clinical judgment, updated psychosocial assessment, and personalized decision-making based on nuanced client status changes that AI cannot reliably perform end-to-end today. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Treatment plan modification is typically a legally and ethically regulated clinical function that must be performed or signed by a licensed social worker or healthcare provider; liability, patient safety, and regulatory requirements create hard barriers to automation. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Treatment plan modifications typically require licensed professional judgment and sign-off, with liability and regulatory documentation standards limiting full automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI inference for drafting suggestions might be low-cost, but the total cost of AI integration, validation, and oversight by a human clinician would likely exceed the cost of the social worker performing the task directly. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | AI could cheaply draft plan updates, but human review, judgment, and liability mean the human cost cannot be substantially avoided, keeping overall cost comparable. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product performs independent treatment plan modification in healthcare; this remains a human-centered clinical function requiring licensed oversight and accountability for client outcomes. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed products autonomously modify social work treatment plans; this remains a clinician-driven task with AI only assisting documentation. |
Organize support groups or counsel family members to assist them in understanding, dealing with, and supporting the client or patient.
7CI 4–11 · exposure 0 · augmentation 38 · importance 4.1/5 · click for rater detail
Organize support groups or counsel family members to assist them in understanding, dealing with, and supporting the client or patient.
7| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare and social services are among the slowest sectors to automate human-facing clinical roles. Regulatory requirements, liability concerns, and strong professional and organizational resistance to replacing clinical judgment limit practical adoption. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare social work is a relationship-driven, in-person field with historically slow AI adoption for direct client/family counseling, though administrative-adjacent AI tools are spreading slowly. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could marginally assist by drafting psychoeducational materials or summarizing family dynamics notes, but it offers little to enhance the core skill of counseling itself—deep listening, emotional attunement, and real-time therapeutic response remain distinctly human. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help draft psychoeducational materials, summarize case notes, suggest discussion topics for support groups, or provide reference resources, meaningfully aiding preparation even though the core counseling remains human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | This task requires genuine emotional intelligence, adaptive counseling based on individual family dynamics, and the ability to respond to unpredictable human vulnerabilities. Current AI systems cannot reliably conduct therapeutic counseling or facilitate group support dynamics with the nuance and ethical responsibility this work demands. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires live emotional attunement, trust-building, and clinical judgment with vulnerable family members that current AI cannot replicate end-to-end; no off-the-shelf system can autonomously run this counseling task at equal quality. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Significant legal and regulatory barriers protect this work: social workers must be licensed in most jurisdictions, therapeutic relationships carry strict standards of care, and liability for poor counseling outcomes falls on the responsible practitioner. Human licensure is a hard legal requirement. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Counseling family members in healthcare contexts typically requires licensure, ethical accountability, and human-contact expectations, plus liability concerns around emotionally sensitive guidance, creating strong professional and regulatory barriers. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 2/5 | AI inference costs are low, but the overhead of human clinical oversight, liability insurance, and the risk of harm from unsupervised AI far exceeds the cost of direct human counseling in real healthcare settings. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI inference is cheap, any viable use requires heavy human oversight, liability management, and clinical supervision, largely offsetting cost savings versus a trained social worker's wage for this relational task. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product reliably performs therapeutic counseling or support group facilitation at production scale. Chatbots exist but lack the clinical judgment, liability framework, and human connection that families in crisis require and expect. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently organizes support groups or counsels families of patients in production; existing chatbot mental-health tools are narrow, unsupervised, and not integrated into healthcare social work practice for this purpose. |
Develop or advise on social policy and assist in community development.
6CI 5–7 · exposure 0 · augmentation 50 · importance 3.1/5 · click for rater detail
Develop or advise on social policy and assist in community development.
6| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare and social services sectors are among the slowest to adopt automation, particularly for high-judgment advisory tasks. Community development work is deeply embedded in local organizations and long-term relationships, making displacement unlikely in the near term. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Social work and public policy sectors are slow AI adopters generally, with pilots focused on documentation/administrative support rather than policy development or community organizing. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist by synthesizing research, identifying policy precedents, modeling outcomes of policy options, and summarizing community data. However, the task's core work—stakeholder engagement, judgment, and political strategy—remains human-driven, making augmentation meaningful but not transformative. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help draft policy briefs, summarize research, analyze community data, and organize information, providing meaningful support even though the core advisory and relational work remains human-driven. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Developing and advising on social policy requires deep understanding of community needs, stakeholder engagement, political feasibility, and nuanced judgment about trade-offs. Current AI cannot autonomously conduct community needs assessment, negotiate policy priorities, or author defensible policy recommendations at the quality a human expert would produce. |
| Task automatability | claude-sonnet-5 | 1/5 | This task requires synthesizing community needs, stakeholder relationships, political judgment, and advocacy—qualitative human-centered work that AI cannot execute end-to-end even with heavy setup. |
| Adoption barriers | claude-haiku-4-5-20251001 | 4/5 | Social policy and community development work often requires licensed social workers, organizational accountability for policy advice, and stakeholder trust. Legal and professional liability for policy recommendations creates substantial barriers to full automation, and organizations face reputational risk delegating such work to AI. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Policy advising and community engagement typically require professional credentials, institutional trust, accountability, and human relationship-building that create strong organizational and social barriers to automation. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | This task requires specialized expertise (MSW, policy knowledge, community relationships) that commands high wages. AI assistance does not yet reduce the need for human social workers to lead policy development, making AI more expensive relative to hiring qualified staff. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this function, so cost comparison favors the human entirely; any AI use is supplementary research support, not replacement. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product reliably performs social policy development or community development advising end-to-end. While AI can assist with research and data synthesis, the core task of advising decision-makers on policy strategy remains beyond production-ready AI capability. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product develops or advises on social policy or drives community development initiatives autonomously; this remains firmly human-led strategic and relational work. |
Supervise and direct other workers providing services to clients or patients.
4CI 0–7 · exposure 0 · augmentation 38 · importance 3.7/5 · click for rater detail
Supervise and direct other workers providing services to clients or patients.
4| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare and social services remain heavily dependent on human supervisory chains and union agreements; digitization of supervision is nascent. No sector data shows meaningful AI-driven worker supervision in production. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare and social services sectors show slower, more cautious AI adoption for managerial and interpersonal functions compared to administrative or documentation tasks. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could assist supervisors with scheduling, data aggregation, or flagging performance outliers, but the core task of directing workers remains human-dependent. Augmentation potential is limited to narrow administrative support. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI tools can help with scheduling, performance tracking, documentation review, and flagging issues to support a supervisor, but the core directing and mentoring role remains human-driven. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Supervising and directing workers requires real-time judgment, interpersonal communication, performance feedback, and adaptive coaching—tasks that demand human emotional intelligence and contextual decision-making. Current AI systems cannot reliably replace a supervisor's ongoing direction, accountability, and personnel management. |
| Task automatability | claude-sonnet-5 | 1/5 | Supervising and directing human workers requires interpersonal leadership, judgment about staff performance, and real-time coordination that current AI cannot execute end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Labor law, employment regulations, and organizational hierarchy require a licensed human manager to hire, direct, and hold workers accountable. Liability and employment law create hard legal barriers to delegating supervisory authority to AI systems. |
| Adoption barriers | claude-sonnet-5 | 4/5 | Supervisory roles in healthcare social work often require licensure, accountability for clinical/ethical oversight, and legal responsibility that cannot be delegated to a non-human system. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | Implementing AI supervision would require significant infrastructure, oversight, and legal liability frameworks that exceed the cost of human supervisors. The human cost of errors in worker direction is high, making AI more expensive when risk is priced in. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | There is no viable AI substitute performing this managerial task, so cost comparison favors the human by default. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product supervises human workers in production settings today. While AI can assist with scheduling or documentation, the supervisory task itself—evaluating performance, making staffing decisions, providing corrective direction—requires human authority and accountability. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product performs direct supervisory management of healthcare social work staff; this remains a human management function. |
Counsel clients and patients in individual and group sessions to help them overcome dependencies, recover from illness, and adjust to life.
3CI 0–6 · exposure 0 · augmentation 38 · importance 4.2/5 · click for rater detail
Counsel clients and patients in individual and group sessions to help them overcome dependencies, recover from illness, and adjust to life.
3| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare and mental health sectors remain heavily human-dependent and risk-averse; therapeutic work is not being displaced by AI in production, and regulatory and professional culture strongly protect the human clinician role. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare and social services sectors show cautious, slow AI adoption for direct clinical/therapeutic tasks, with pilots emerging mainly for documentation and triage support rather than counseling itself. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI may assist with scheduling, documentation, or psychoeducational materials, but the core therapeutic interaction—listening, responding to emotion, and guiding recovery—remains fundamentally human; augmentation is minimal on the counseling task itself. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with session prep, treatment plan drafting, resource matching, note-taking, and follow-up reminders, meaningfully aiding the social worker's workflow without replacing the core counseling interaction. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Counseling clients to overcome dependencies and adjust to life requires deep empathy, nuanced judgment, therapeutic relationship-building, and real-time adaptation to emotional and psychological states—none of which current AI systems can reliably replicate end-to-end at clinical quality or with meaningful time savings over a human clinician. |
| Task automatability | claude-sonnet-5 | 1/5 | Therapeutic counseling requires real-time empathy, trust-building, ethical judgment, and adaptive human interaction that current AI cannot replicate end-to-end at equal quality for vulnerable clinical populations. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Counseling and therapeutic work is legally and ethically regulated; a licensed clinical social worker must be directly responsible for client care, diagnosis, and treatment planning in virtually all jurisdictions, creating a hard barrier to full automation. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Clinical counseling for healthcare social work is subject to licensure, confidentiality law (HIPAA), liability for patient harm, and ethical mandates requiring a qualified human professional to conduct and be accountable for sessions. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI cannot perform this task, so no cost comparison applies; where AI might assist (e.g., intake screening chatbots), the integration, human oversight, and liability costs exceed savings from a licensed social worker's perspective. |
| Cost vs. human wage | claude-sonnet-5 | 2/5 | While AI inference is cheap, the lack of a viable automated substitute means comparing cost is largely moot; any AI-assisted tools still require licensed clinician oversight, keeping effective costs comparable to or higher than pure automation savings. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product reliably performs therapeutic counseling for addiction recovery or mental health adjustment; such work requires licensure, clinical judgment, and legal/ethical accountability that cannot be substituted by current AI systems in any production setting. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed product independently conducts clinical counseling sessions for illness recovery or dependency treatment; AI chatbots exist as supplementary tools but not as substitutes performing this task reliably in production. |
Advocate for clients or patients to resolve crises.
1CI 0–3 · exposure 0 · augmentation 38 · importance 4.7/5 · click for rater detail
Advocate for clients or patients to resolve crises.
1| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Healthcare social work is a regulated, human-centered profession with low AI adoption for core advocacy tasks. Organizational and regulatory structures strongly reinforce human social workers as the required legal and ethical actor, limiting experimentation with automation. |
| Sector adoption velocity | claude-sonnet-5 | 2/5 | Healthcare social services broadly show slow, cautious AI adoption due to high-stakes human contact, privacy regulation, and lack of mature advocacy-specific tools. |
| Augmentation potential | claude-haiku-4-5-20251001 | 2/5 | AI could provide marginal assistance by drafting letters, organizing case information, or suggesting advocacy strategies for the social worker to review, but the core work—crisis resolution and client advocacy—remains human-driven and AI offers limited augmentation of the actual advocacy act itself. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can help draft case notes, research resources, or summarize patient records to prepare for advocacy, but cannot handle the interpersonal and institutional negotiation itself. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Advocating for clients to resolve crises requires nuanced understanding of individual circumstances, emotional intelligence, negotiation with third parties, and judgment about when to escalate or de-escalate. Current AI cannot reliably perform the interpersonal negotiation, crisis assessment, and contextual decision-making needed to resolve crises on behalf of vulnerable populations. |
| Task automatability | claude-sonnet-5 | 1/5 | Crisis advocacy requires real-time judgment, negotiation with institutions, emotional attunement, and trust-building that current AI cannot perform end-to-end; no plausible 50% time-saving at equal quality exists. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | This task is heavily protected by licensing (social work credentials required in most jurisdictions), liability (errors in crisis advocacy can cause serious harm), duty-of-care requirements, and regulatory mandates that a qualified human social worker must assess and act on behalf of vulnerable clients. Crisis intervention has explicit human-contact requirements in virtually all regulatory frameworks. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Healthcare social work advocacy is tied to licensure, ethical codes, confidentiality, and legal representation duties, requiring a credentialed human to act and be accountable. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | AI systems cannot currently perform this task end-to-end, so direct cost comparison is not applicable; the human remains necessary and AI cannot reduce cost-per-outcome for crisis advocacy. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the human function here, so any 'cost' comparison favors the human entirely; there is no viable AI-only cost baseline. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed AI product can autonomously advocate for a patient or client in crisis resolution, which requires real-time interaction with institutional actors, legal/medical authority, and crisis judgment. While chatbots can provide information, they cannot meaningfully negotiate with hospitals, courts, or services to resolve actual crises. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product performs crisis advocacy on behalf of patients; this remains firmly in the human relational and institutional domain. |
Investigate child abuse or neglect cases and take authorized protective action when necessary.
0CI 0–0 · exposure 0 · augmentation 50 · importance 4.5/5 · click for rater detail
Investigate child abuse or neglect cases and take authorized protective action when necessary.
0| Dimension | Model | Rating | Rationale |
|---|---|---|---|
| Sector adoption velocity | claude-haiku-4-5-20251001 | 1/5 | Adoption of AI for investigations or protective decisions in child welfare is minimal; the sector remains heavily regulated, human-centered, and conservative due to high stakes, with limited digitization and strong professional gatekeeping. |
| Sector adoption velocity | claude-sonnet-5 | 1/5 | Child welfare and protective services is a highly regulated, in-person, government-run sector with minimal AI adoption for core casework decisions. |
| Augmentation potential | claude-haiku-4-5-20251001 | 3/5 | AI can assist with case documentation, flagging historical records, organizing evidence, and literature research, improving worker efficiency in preparation and analysis, but the core investigative and protective judgment remains human-centered. |
| Augmentation potential | claude-sonnet-5 | 3/5 | AI can assist with documentation, case note drafting, risk-scoring flags, and record review, but the core investigative and protective actions remain human-led. |
| Task automatability | claude-haiku-4-5-20251001 | 1/5 | Investigating abuse/neglect requires nuanced interviews with vulnerable populations, interpretation of behavioral and contextual cues, judgment about credibility and imminent danger, and legal/ethical decision-making that demands human expertise and accountability. No AI system can reliably perform the investigative work end-to-end today. |
| Task automatability | claude-sonnet-5 | 1/5 | This requires physical home visits, in-person interviews, judgment about safety, and legal authority to remove children—none of which current AI can perform end-to-end. |
| Adoption barriers | claude-haiku-4-5-20251001 | 5/5 | Substantial legal and regulatory barriers exist: only licensed social workers have authority to conduct investigations and authorize protective actions; liability for erroneous conclusions is borne by the professional; child protection is a mandated human-responsibility framework with statutory requirements. |
| Adoption barriers | claude-sonnet-5 | 5/5 | Child protective investigations require legally authorized, licensed social workers with statutory authority to intervene, and errors carry severe legal and safety consequences. |
| Cost vs. human wage | claude-haiku-4-5-20251001 | 1/5 | The cost of an AI system plus required human oversight, verification, and liability mitigation far exceeds the loaded wage of a trained social worker, given the high stakes and mandatory human sign-off on all material decisions. |
| Cost vs. human wage | claude-sonnet-5 | 1/5 | AI cannot substitute for the human labor involved, so there is no viable cost comparison—human investigators remain mandatory. |
| Technical feasibility today | claude-haiku-4-5-20251001 | 1/5 | No deployed product performs child abuse investigations or authorized protective actions independently. While AI assists with documentation and data analysis, the investigatory and protective-action components require licensed professionals with legal authority and liability exposure that precludes AI substitution. |
| Technical feasibility today | claude-sonnet-5 | 1/5 | No deployed AI product conducts child abuse investigations or executes protective custody actions; this remains entirely a human function. |
Related occupations — Community & Social Service
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.