Mental Health and Substance Abuse Social Workers

21-1023.00
Median wage $60,280/yr132,810 employed (US)Rank #742 of 923 scored · top 80% by substitution

Assess and treat individuals with mental, emotional, or substance abuse problems, including abuse of alcohol, tobacco, and/or other drugs. Activities may include individual and group therapy, crisis intervention, case management, client advocacy, prevention, and education.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution17
Exposure14
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

13 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%12

panel mean rating 1.5/5 → substitution pressure 12/100

Technical feasibility todayw 20%16

panel mean rating 1.7/5 → substitution pressure 16/100

Cost vs. human wagew 15%21

panel mean rating 1.9/5 → substitution pressure 21/100

Adoption barriersw 20%inverted — strong barriers lower the score20

panel mean rating 4.2/5 (barrier strength) → substitution pressure 20/100

Sector adoption velocityw 10%22

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

Task breakdown (13 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.

Increase social work knowledge by reviewing current literature, conducting social research, or attending seminars, training workshops, or classes.

44

CI 39–50 · exposure 30 · augmentation 75 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Social work is a mid-digitization sector where some organizations use AI-assisted literature tools, but widespread production adoption of AI for professional development remains inconsistent. Many agencies still rely on traditional in-person training and manual reading.
Sector adoption velocityclaude-sonnet-52/5Social work and mental health sectors show slower AI adoption overall compared to information/finance sectors, with pilots for AI-assisted research tools but limited institutional deployment.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can substantially accelerate the efficiency of literature review, automatically curate relevant papers, summarize findings, and flag emerging best practices, allowing workers to focus their time on deeper analysis and application to their practice. This assistive role is already valuable in production.
Augmentation potentialclaude-sonnet-54/5AI tools (e.g., literature summarizers, research assistants, personalized learning recommendations) can meaningfully speed up how social workers review literature and prepare for continuing education, even though attendance-based learning remains human-driven.
Task automatabilityclaude-haiku-4-5-202510012/5AI can assist with literature review and summarizing research, but cannot independently attend seminars or workshops, and genuine learning acquisition requires human comprehension and integration. The task involves personal professional development that benefits from human judgment about relevance and gaps in knowledge.
Task automatabilityclaude-sonnet-52/5AI can help summarize literature or research, but 'increasing knowledge' as a professional developmental task also involves attending live seminars/workshops and internalizing learning, which cannot be fully automated end-to-end.'
Adoption barriersclaude-haiku-4-5-202510012/5There are minimal legal or regulatory barriers to using AI to assist with knowledge acquisition; professional standards encourage continuing education but do not mandate a specific method of staying current. Organizational culture may prefer human-led training but does not legally require it.
Adoption barriersclaude-sonnet-52/5No strict licensing requirement mandates the literature review itself be done by a human, though continuing education credits often require verified attendance or engagement, creating mild structural friction.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI-assisted literature review and training-module summaries cost a fraction of a human's time spent on the same activities; however, workshop attendance and real-time interactive learning still require human presence, so full task substitution is incomplete.
Cost vs. human wageclaude-sonnet-53/5AI-assisted literature review/summarization tools are cheap relative to human time spent reading, but the task also includes attending workshops/seminars, which AI doesn't replace, keeping overall cost savings moderate.
Technical feasibility todayclaude-haiku-4-5-202510013/5AI systems today can summarize papers, extract key findings, and generate reading lists with reasonable accuracy, but products do so with material limitations in domain specificity and nuance for specialized social work literature. Human review of AI-generated summaries remains necessary.
Technical feasibility todayclaude-sonnet-52/5AI literature summarization and research tools exist and are used somewhat, but no deployed product autonomously performs continuing professional education or research synthesis reliably for social workers at scale.

Educate clients or community members about mental or physical illness, abuse, medication, or available community resources.

31

CI 29–34 · exposure 30 · augmentation 75 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Mental health and substance abuse organizations operate in heavily regulated, relationship-intensive contexts with slow digital adoption relative to information sectors. Most agencies still rely on in-person or phone education; AI-driven automated education is rare in production within this sector.
Sector adoption velocityclaude-sonnet-52/5Social services and healthcare-adjacent behavioral health settings are slower AI adopters than corporate/professional services sectors, with pilots emerging but production use still limited.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can meaningfully assist social workers by drafting educational materials, generating multilingual resources, organizing community resource databases, and creating visual aids—all of which reduce preparation time and allow workers to focus on tailoring messages and building rapport with clients.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully help social workers by drafting educational handouts, summarizing resources, and personalizing information for literacy or language needs, greatly speeding preparation while the worker still delivers it.
Task automatabilityclaude-haiku-4-5-202510012/5AI can generate educational content and materials, but the task requires responsiveness to individual client circumstances, emotional engagement, and adaptation to diverse comprehension levels. Current systems cannot reliably perform the full task end-to-end with 50% time savings at equal quality because effective patient education depends on understanding the specific client's readiness, trauma history, and cultural context.
Task automatabilityclaude-sonnet-52/5AI can generate accurate educational content about illness, medication, or resources, but effective delivery requires reading the client's emotional state, tailoring language to trauma history, and building trust, which current systems cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5High barriers exist due to legal liability (social workers hold professional licensure and duty of care), regulatory requirements around healthcare information, and organizational/ethical expectations that vulnerable clients receive human connection and accountability. Automation of this task carries meaningful malpractice and consent risks.
Adoption barriersclaude-sonnet-54/5Work with mental health/substance abuse clients often involves mandated reporting, licensure requirements, confidentiality law, and liability concerns that require a credentialed human to be involved in client education and disclosure.
Cost vs. human wageclaude-haiku-4-5-202510013/5AI-generated educational content and delivery systems (chatbots, video platforms) can reduce per-contact labor costs, but require substantial oversight, clinical vetting, and customization, making the total-cost comparison with a trained social worker roughly equivalent once all integration and quality control are included.
Cost vs. human wageclaude-sonnet-53/5AI-generated educational materials or chatbot interactions are far cheaper per interaction than a social worker's time, but oversight, liability review, and quality control for sensitive populations narrow the savings.
Technical feasibility todayclaude-haiku-4-5-202510013/5Educational chatbots and content generation tools exist and are deployed in some healthcare settings, but they lack the reliability and human connection necessary for vulnerability-sensitive populations. Real-world systems achieve partial automation (e.g., distributing pre-written materials) but struggle with personalization and handling sensitive disclosures.
Technical feasibility todayclaude-sonnet-52/5Chatbots and AI health-information tools exist and are used for psychoeducation, but they are not reliably deployed as substitutes for licensed social workers educating vulnerable clients on sensitive topics like abuse or substance use.

Assist clients in adhering to treatment plans, such as setting up appointments, arranging for transportation to appointments, or providing support.

28

CI 25–30 · exposure 25 · augmentation 50 · importance 4.0/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Mental health and social services sectors have historically slow AI adoption due to regulatory constraints, client vulnerability, and preference for human continuity of care. While text-based support tools exist, organizations are cautious and adoption remains limited to pilot and supplementary roles.
Sector adoption velocityclaude-sonnet-52/5Social work and behavioral health remain low-digitization, resource-constrained sectors with slow AI adoption despite growing interest in patient engagement tools.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can usefully assist by automating appointment scheduling, sending reminders, flagging missed sessions, and providing initial information, raising a human social worker's capacity to focus on engagement and crisis response. However, the core support function remains human-centric and augmentation is partial rather than transformative.
Augmentation potentialclaude-sonnet-53/5AI can meaningfully assist with appointment reminders, scheduling logistics, and even resource look-up for transportation options, improving efficiency while the social worker retains the support role.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can handle appointment scheduling and transportation logistics (discrete, structured subtasks), the core responsibility—providing support and building trust with vulnerable clients—requires human empathy and contextual judgment. Current AI cannot reliably perform the emotional scaffolding or adaptive encouragement that constitutes the bulk of this task.
Task automatabilityclaude-sonnet-52/5Scheduling and reminder components could be automated, but arranging transportation and 'providing support' require relational, judgment-based, and logistical human interaction that current AI cannot fully replicate end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Significant regulatory and licensing barriers apply: mental health and substance abuse treatment involves liability, client confidentiality (HIPAA), and in many jurisdictions legally mandated human contact and clinical judgment. Organizations and insurers require licensed practitioners for treatment accountability.
Adoption barriersclaude-sonnet-53/5No licensing requirement blocks scheduling automation, but liability concerns, client vulnerability, and the need for a trusted human relationship in support-giving create moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI could automate appointment scheduling cheaply, but the supervision, quality assurance, and human social worker oversight required for adherence support (including liability and re-engagement after failure) keep total cost per outcome near or above human wages.
Cost vs. human wageclaude-sonnet-52/5AI scheduling tools are cheap, but the human support and care-coordination portions still require paid staff, so overall cost savings are limited when the full task is considered.
Technical feasibility todayclaude-haiku-4-5-202510012/5Chatbots and scheduling tools exist and perform narrow components (booking, reminders), but no deployed product reliably provides holistic treatment-adherence support with the personalization, crisis recognition, and relationship continuity this role demands. Integration remains research-stage for the full task.
Technical feasibility todayclaude-sonnet-52/5Scheduling assistants and reminder chatbots exist in some clinics, but comprehensive coordination of transportation and emotional support for vulnerable populations is not reliably handled by deployed AI products.

Monitor, evaluate, and record client progress with respect to treatment goals.

25

CI 25–25 · exposure 25 · augmentation 63 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI in mental health and substance abuse sectors remains slow; most social work agencies operate in under-resourced, risk-averse environments with limited digitization and high regulation. Pilot projects exist but production displacement is minimal compared to information-sector adoption patterns.
Sector adoption velocityclaude-sonnet-52/5Social work and behavioral health remain lower-digitization sectors with slow, cautious AI adoption due to privacy, regulatory, and trust concerns.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by helping organize and flag trends in structured data, summarizing notes, and surfacing metrics that aid the social worker's review process. However, augmentation is limited because the core evaluative judgment—connecting progress to goals and adjusting treatment—remains fundamentally human.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist by drafting progress notes, flagging patterns in client data, and organizing documentation, improving efficiency while the clinician retains judgment and oversight.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with data recording and pattern detection in structured metrics, the task fundamentally requires human clinical judgment to interpret nuanced behavioral and emotional progress against individualized treatment goals. Current AI cannot reliably perform the evaluative and monitoring aspects that depend on understanding context, rapport, and subtle client indicators at ≥50% time savings with equal quality.
Task automatabilityclaude-sonnet-52/5AI can help summarize notes and track structured data, but clinical judgment about progress toward psychosocial treatment goals requires nuanced human observation and interpretation that isn't fully replaceable end-to-end today.
Adoption barriersclaude-haiku-4-5-202510014/5Regulatory and liability barriers are substantial: documentation must meet legal/clinical standards, social workers bear responsibility for treatment decisions, and most jurisdictions require licensed professionals to own treatment evaluation and progress assessment. Liability asymmetry—a missed deterioration is high-cost—further protects the human role.
Adoption barriersclaude-sonnet-54/5Licensed social workers are legally and ethically required to make and document clinical judgments about client progress, and liability/confidentiality concerns strongly limit full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Integration of AI tooling into social work workflows requires clinical oversight, data governance, and human review to ensure safety and accuracy. The cost of AI infrastructure, data management, and necessary human validation approaches or exceeds the incremental efficiency gain over a social worker's direct documentation.
Cost vs. human wageclaude-sonnet-52/5AI documentation assistants can reduce time on notes, but human clinician evaluation and judgment remain necessary, keeping overall cost comparable to human-only workflows.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some products exist for automated clinical note generation and simple metric tracking, but they have material limitations in capturing qualitative progress, handling complex co-morbidities, and reliably assessing treatment efficacy. No mature deployed system reliably performs the full monitoring and evaluation task independently at production scale.
Technical feasibility todayclaude-sonnet-52/5Some EHR-integrated tools offer note summarization and sentiment/progress tracking, but no deployed product reliably evaluates clinical progress against individualized treatment goals at scale.

Refer patient, client, or family to community resources for housing or treatment to assist in recovery from mental or physical illness, following through to ensure service efficacy.

25

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

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Social service organizations are typically under-digitized, risk-averse, and heavily regulated; pilot-stage adoption of AI assistants exists but production displacement remains minimal and slow.
Sector adoption velocityclaude-sonnet-52/5Social work and behavioral health remain relatively low-tech and slow to adopt AI systems compared to sectors like finance or information services, with most current use limited to administrative support.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by automating resource database searches, generating matching suggestions, and scheduling follow-ups, meaningfully reducing administrative burden while the worker retains judgment and relationship responsibility.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist by surfacing relevant local resources, drafting referral documentation, and flagging follow-up reminders, improving efficiency while the social worker retains responsibility for judgment and relationship management.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can identify and retrieve resource information, the task requires nuanced judgment about individual needs, follow-up communication, and adaptive problem-solving when services fail—capabilities that current systems cannot reliably manage end-to-end at quality parity with humans.
Task automatabilityclaude-sonnet-52/5AI can help identify and compile relevant community resources quickly, but the judgment-heavy matching, relationship-building, and follow-through verification of service efficacy require human involvement and cannot be fully automated today.
Adoption barriersclaude-haiku-4-5-202510014/5Liability and professional responsibility create strong barriers: social workers are legally and ethically accountable for referral quality and follow-through; clients often need human trust and advocacy; organizations typically require licensed staff to own the decision.
Adoption barriersclaude-sonnet-54/5This task involves clinical judgment, confidentiality obligations, and licensure requirements for mental health social workers, plus liability concerns around ensuring appropriate care continuity, creating strong barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI can reduce overhead on routine lookups and initial contact, but the labor cost of human judgment, relationship management, and accountability still dominates; integration and oversight costs are non-trivial relative to modest displacement.
Cost vs. human wageclaude-sonnet-52/5AI tools could reduce time spent searching for resources, but the follow-up, relationship management, and liability-sensitive coordination still require paid professional time, keeping overall cost savings modest.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some products can generate resource lists or send referral notifications, but deployed systems lack reliable capability to assess fit, manage relationship continuity, handle exceptions, and verify service efficacy through sustained follow-up.
Technical feasibility todayclaude-sonnet-52/5Some resource-navigation chatbots and case management tools exist, but they are narrow in scope and not widely deployed to reliably handle the full referral-and-follow-up cycle in clinical social work settings.

Plan or conduct programs to prevent substance abuse, combat social problems, or improve health or counseling services in community.

18

CI 11–25 · exposure 13 · augmentation 50 · importance 3.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Social work and community health sectors lag in AI adoption relative to finance or tech; most organizations still rely on manual program planning, and AI-assisted tools remain in pilot stages rather than production deployment. Resource constraints in nonprofits and public agencies slow adoption velocity.
Sector adoption velocityclaude-sonnet-52/5Social services and community health sectors are generally slow adopters of AI tools relative to information/finance industries, with pilots more common than production use.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can meaningfully assist by generating evidence summaries, identifying intervention models from literature, analyzing community health data, or drafting program proposals—productivity gains for human planners. However, the scope is limited to information synthesis; final program design and community engagement remain firmly human-driven.
Augmentation potentialclaude-sonnet-53/5AI can help with needs assessments, literature reviews, grant writing, and drafting program materials, meaningfully assisting planning even though humans must lead design and implementation.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with program planning (generating frameworks, identifying resources) and data analysis for needs assessment, the core task requires human judgment on community priorities, stakeholder engagement, and culturally-sensitive design. End-to-end automation would miss critical interpersonal and contextual elements essential to effective prevention and counseling programs.
Task automatabilityclaude-sonnet-51/5This is a high-level, judgment-heavy program design and community engagement task requiring stakeholder relationships, local context knowledge, and strategic planning that current AI cannot execute end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Social work practice is governed by professional licensing, ethical codes, and liability frameworks that require a licensed human to bear responsibility for program design and clinical judgment. Community trust and legal accountability for outcomes further mandate human professional involvement in program planning and service delivery.
Adoption barriersclaude-sonnet-54/5Social work licensure, ethical accountability, community trust, and human-contact requirements for counseling and outreach create strong barriers to full automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for data analysis and program drafting cost less per unit task, but a loaded human social worker's effective cost for full program planning and execution is still lower overall when accounting for the extensive oversight, redesign, and relationship-building AI solutions require.
Cost vs. human wageclaude-sonnet-52/5AI can cheaply produce drafts or research summaries, but the actual planning, coalition-building, and program execution still require costly human labor, keeping overall cost comparable to or above human-only delivery.
Technical feasibility todayclaude-haiku-4-5-202510012/5No deployed system reliably conducts or plans comprehensive substance-abuse prevention programs autonomously. AI can draft program proposals, analyze outcomes data, or suggest evidence-based interventions, but real-world program execution requires human social workers to navigate community relationships, adapt to local needs, and oversee implementation—tasks beyond current AI capabilities.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product plans or runs community-level prevention/counseling programs autonomously; this remains firmly in the human professional domain.

Modify treatment plans according to changes in client status.

14

CI 3–25 · exposure 13 · augmentation 63 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Mental health and substance abuse services remain heavily regulated and human-centered; even digitization is slow in many organizations. Adoption of autonomous AI for clinical decision-making is minimal, with most sectors still in pilot or early adoption phases for assistive tools.
Sector adoption velocityclaude-sonnet-52/5Healthcare and behavioral health sectors show slower, more cautious AI adoption for clinical decision-making due to regulatory and safety concerns, with pilots more common than production deployment.
Augmentation potentialclaude-haiku-4-5-202510013/5AI could assist by summarizing client status changes, suggesting evidence-based interventions, or flagging inconsistencies in progress notes, thereby reducing documentation burden and freeing the clinician for higher-level judgment. However, the core modification task remains human-driven.
Augmentation potentialclaude-sonnet-54/5AI can help summarize client progress notes, flag status changes, and suggest evidence-based treatment adjustments for clinician review, meaningfully speeding up plan updates while the clinician retains final judgment.
Task automatabilityclaude-haiku-4-5-202510011/5Modifying treatment plans requires nuanced clinical judgment, understanding of individual client circumstances, and integration of new information into a complex care framework. Current AI cannot reliably make clinical decisions that affect mental health outcomes or adjust therapeutic approaches without human oversight, falling well short of 50% time saving at equal quality.
Task automatabilityclaude-sonnet-52/5Requires clinical judgment based on nuanced client interactions, risk assessment, and therapeutic relationship context that current AI cannot reliably interpret or act on end-to-end.itting.
Adoption barriersclaude-haiku-4-5-202510014/5Treatment plan modification is typically the legal and professional responsibility of licensed social workers and other credentialed clinicians. Regulatory frameworks in mental health and substance abuse treatment require human sign-off, and liability for inadequate plan adjustments falls on the licensed practitioner, creating strong protective barriers against full automation.
Adoption barriersclaude-sonnet-55/5Treatment plan modifications typically require a licensed social worker or clinician's professional judgment and signature, with strong regulatory, ethical, and liability constraints.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for documentation or data flagging may reduce administrative overhead, but the clinical work of plan modification still requires a licensed professional's expertise, meaning total cost-in-use remains comparable to or higher than manual processes given integration and oversight needs.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the clinician's judgment and liability role, so the effective cost comparison favors the human since AI alone cannot produce an equivalent, legally valid output.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can flag changes in client status data or suggest template language, no deployed product reliably modifies clinical treatment plans independently. Existing systems lack the contextual understanding of therapeutic relationships and client-specific factors needed for autonomous plan modification in mental health care.
Technical feasibility todayclaude-sonnet-51/5No deployed products autonomously modify clinical treatment plans for mental health/substance abuse clients; this remains a licensed clinician responsibility with only research-stage decision-support tools.

Interview clients, review records, conduct assessments, or confer with other professionals to evaluate the mental or physical condition of clients or patients.

11

CI 3–20 · exposure 13 · augmentation 63 · importance 4.3/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Healthcare and social services sectors show cautious, pilot-stage adoption of AI for screening and documentation aids, but production replacement of clinical assessment is rare and slow due to regulatory, liability, and organizational resistance.
Sector adoption velocityclaude-sonnet-52/5Healthcare and social services, especially behavioral health, show slow, cautious AI adoption for clinical tasks due to regulation, liability, and the sensitive nature of client interactions.
Augmentation potentialclaude-haiku-4-5-202510014/5Current AI systems substantially assist social workers by summarizing client records, flagging risk factors, suggesting screening questions, and auto-generating documentation drafts—meaningfully raising productivity while the clinician retains assessment authority and decision-making.
Augmentation potentialclaude-sonnet-53/5AI can help by transcribing/summarizing session notes, flagging risk indicators from records, and drafting assessment documentation, but the interview and judgment remain human-led.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can assist with record review and generate initial assessment summaries, the core task—evaluating client mental/physical condition through interviews and clinical judgment—requires human presence, rapport-building, and real-time responsiveness that current AI cannot reliably replicate end-to-end. Significant setup and human oversight would be needed for any meaningful automation.
Task automatabilityclaude-sonnet-51/5This requires building rapport, reading nonverbal cues, and making clinical judgments about vulnerable clients—AI cannot conduct the full interview and assessment process end-to-end at equal quality today.
Adoption barriersclaude-haiku-4-5-202510015/5Regulatory and professional licensing requirements are substantial: mental health assessments typically must be conducted and certified by licensed clinical professionals (LCSW, counselor, psychiatrist), and liability for misdiagnosis or harm falls on the responsible clinician. Legal and ethical frameworks mandate human accountability.
Adoption barriersclaude-sonnet-55/5Clinical assessment of mental/physical condition typically requires a licensed social worker or clinician, involves liability for missed diagnoses (e.g., suicide risk), and mandates human accountability and sign-off.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI tools for record analysis and note generation are relatively inexpensive, but the irreducible human labor (licensed assessment, clinical judgment, legal liability) means overall cost per complete, defensible evaluation remains comparable to or higher than a human social worker performing the task.
Cost vs. human wageclaude-sonnet-51/5Since AI cannot substitute for the licensed professional performing this task, there is no viable AI-only cost comparison; any AI use is supplementary, not cost-substitutive.
Technical feasibility todayclaude-haiku-4-5-202510012/5Some deployed systems exist for initial screening questions and record summarization, but no production system reliably conducts full clinical interviews or produces defensible assessments without human oversight. Error rates and liability concerns remain high.
Technical feasibility todayclaude-sonnet-51/5No deployed product performs independent clinical mental health assessments and interviews of patients in production; this remains firmly a human clinical function.

Develop or advise on social policy or assist in community development.

11

CI 11–11 · exposure 0 · augmentation 63 · importance 2.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Social services and nonprofit sectors show slower AI adoption overall; policy development remains a domain where human expertise and trust are paramount. Pilot use of AI for research support may occur, but production-level replacement is minimal and adoption remains largely in early stages.
Sector adoption velocityclaude-sonnet-52/5Social work and community organizing sectors have low overall AI adoption for strategic/advisory functions, with pilots limited to administrative support rather than policy development itself.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist human policy advisors by synthesizing research, identifying data patterns, drafting policy briefs, and suggesting evidence-based options, meaningfully raising their productivity. However, the human expert must remain in control of judgment and community engagement, so augmentation is valuable but not transformative.
Augmentation potentialclaude-sonnet-54/5AI can meaningfully assist by synthesizing research, drafting policy briefs, analyzing community needs data, and summarizing precedents, enhancing a social worker's efficiency while they retain judgment and community engagement roles.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires deep understanding of community needs, stakeholder engagement, political dynamics, and normative judgment about social priorities—all of which demand human expertise and discretionary reasoning that current AI cannot perform end-to-end. AI cannot independently develop or meaningfully advise on social policy without substantial human direction and validation.
Task automatabilityclaude-sonnet-51/5This task requires synthesizing local context, stakeholder relationships, political judgment, and community trust-building that AI cannot perform end-to-end; no current system can substitute for the human role of advising or shaping policy in a community context.
Adoption barriersclaude-haiku-4-5-202510014/5Social policy development and community advising are typically governed by professional licensure (MSW), ethical codes, and liability frameworks that require a licensed human to take responsibility for recommendations. Organizations and communities expect human expertise and accountability, creating both legal and organizational friction against full substitution.
Adoption barriersclaude-sonnet-54/5Social policy advising and community development often require professional licensure, accountability to governing bodies, and trusted human relationships with communities, creating strong organizational and legitimacy barriers to AI substitution.
Cost vs. human wageclaude-haiku-4-5-202510012/5While AI text generation is cheap, the true cost of policy advising includes domain expertise, stakeholder consultation, and accountability for outcomes. Current AI cannot replace these functions; it may only reduce some research legwork, making overall cost savings marginal relative to a human social policy expert's loaded wage.
Cost vs. human wageclaude-sonnet-52/5While AI can cheaply generate background research or draft language, the human labor of stakeholder engagement, negotiation, and community relationship-building dominates cost and cannot be offloaded, keeping overall cost comparable to or only slightly better than fully human effort.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs policy development or community development advising in production contexts. AI tools can assist with research or drafting, but no mature system independently delivers policy recommendations or community strategies that organizations rely on without expert human oversight.
Technical feasibility todayclaude-sonnet-51/5No deployed product independently develops social policy or leads community development efforts; AI is at most a research/drafting aid, not a demonstrated production performer of this task.

Supervise or direct other workers who provide services to clients or patients.

6

CI 0–13 · exposure 5 · augmentation 38 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Mental health and social services sectors remain heavily human-centric, with strong professional norms and regulatory requirements that prioritize licensed clinical supervision. Adoption of AI for actual supervisory direction is negligible.
Sector adoption velocityclaude-sonnet-52/5Social services and healthcare administration sectors are slower adopters of AI-driven management tools compared to finance or tech, with pilots for administrative support but not for supervisory replacement.
Augmentation potentialclaude-haiku-4-5-202510012/5AI can provide useful analytics on staff performance metrics, schedule optimization, and documentation tracking, but these are narrow supports that do not fundamentally transform a supervisor's core task of evaluating, coaching, and directing staff in real time.
Augmentation potentialclaude-sonnet-53/5AI can assist supervisors with administrative tasks like scheduling, documentation review, and performance tracking, but does not replace the interpersonal supervisory function.
Task automatabilityclaude-haiku-4-5-202510011/5Supervising or directing other workers requires real-time judgment, interpersonal feedback, performance calibration, and accountability decisions that depend on human context and emotion. Current AI cannot reliably replace this supervisory function end-to-end.
Task automatabilityclaude-sonnet-51/5Supervising and directing human staff involves judgment, mentorship, accountability, and interpersonal leadership that current AI cannot perform end-to-end.'
Adoption barriersclaude-haiku-4-5-202510015/5Legal and regulatory frameworks in healthcare and social services typically require a licensed human to supervise and direct clinical staff. Liability, accreditation standards, and accountability for client outcomes create hard barriers to AI substitution.
Adoption barriersclaude-sonnet-54/5Clinical supervision in mental health/substance abuse settings often requires licensed, credentialed supervisors per regulatory and accreditation standards, creating strong legal and organizational barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5Supervision requires integrated knowledge of staff, clients, regulations, and organizational culture. AI oversight tools remain supplementary; the labor cost of a human supervisor remains far lower than the total cost of AI infrastructure, integration, and mandatory human oversight fallback.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this supervisory function, so cost comparison favors the human role entirely.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can assist with scheduling, documentation review, and performance metrics, no deployed system reliably supervises or directs human workers in a professional or legal capacity. Some products offer analytics dashboards, but they do not perform the actual supervisory task.
Technical feasibility todayclaude-sonnet-51/5No deployed product manages or directs human clinical/case workers autonomously; AI is at best used for scheduling or documentation support, not supervisory decision-making.

Counsel clients in individual or group sessions to assist them in dealing with substance abuse, mental or physical illness, poverty, unemployment, or physical abuse.

6

CI 0–11 · exposure 5 · augmentation 38 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Mental health and social work remain low-digitization, human-contact-essential sectors with strong professional licensing protections and slow organizational adoption of AI. Regulatory constraints and ethical obligations prevent rapid automation.
Sector adoption velocityclaude-sonnet-52/5Social work and behavioral health are historically slow to adopt AI directly in clinical care due to regulatory, ethical, and trust concerns, though administrative AI tools are spreading faster than clinical ones.
Augmentation potentialclaude-haiku-4-5-202510012/5AI might assist with scheduling, documentation, or intake screening, but current systems offer minimal value in augmenting the core counseling process itself, which requires nuanced human judgment and presence. Limited meaningful assistance on the primary task.
Augmentation potentialclaude-sonnet-53/5AI can assist with session notes, treatment plan drafting, resource matching, and between-session psychoeducational content, meaningfully supporting but not replacing the counselor's core relational work.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires genuine therapeutic relationship-building, real-time emotional attunement, crisis assessment, and adaptive intervention based on individual client needs and disclosure—capabilities that current AI systems cannot reliably perform. AI cannot deliver the human presence, trust, and accountability essential to therapeutic counseling.
Task automatabilityclaude-sonnet-51/5Live therapeutic counseling requires real-time emotional attunement, crisis judgment, and trust-building that current AI cannot replicate end-to-end at equal quality; it is not a task that can be handed off wholesale to AI today.
Adoption barriersclaude-haiku-4-5-202510015/5Strict regulatory requirements (licensure, clinical supervision, duty of care, informed consent) and high legal liability for adverse outcomes legally mandate human clinicians to perform or directly oversee counseling. Patient safety and consent requirements create hard barriers to substitution.
Adoption barriersclaude-sonnet-55/5Clinical counseling for mental health and substance abuse is subject to licensure requirements, mandatory reporting duties, and high liability for harm, all of which legally require a qualified human practitioner.
Cost vs. human wageclaude-haiku-4-5-202510011/5Even accounting for low AI inference costs, the liability exposure, need for human oversight, mandatory supervision by licensed clinicians, and inability to provide real therapeutic value mean the all-in cost per equivalent outcome far exceeds a human social worker's loaded wage.
Cost vs. human wageclaude-sonnet-52/5While inference cost is low, the oversight, liability, and quality-assurance needed to make AI counseling safe raises effective cost, and licensed human labor remains the necessary comparator for most of this task.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs therapeutic counseling for mental health and substance abuse at production scale. Chatbots exist but lack clinical licensure, liability coverage, continuity of care, and demonstrated equivalence to human counseling outcomes.
Technical feasibility todayclaude-sonnet-52/5Some AI chatbots (e.g., Woebot-style apps) offer scripted support, but no deployed product reliably conducts substantive individual or group counseling for substance abuse, abuse, or serious mental illness at production scale.

Counsel or aid family members to assist them in understanding, dealing with, or supporting the client or patient.

6

CI 0–11 · exposure 0 · augmentation 50 · importance 3.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Mental health and substance abuse services are heavily regulated, client-facing, and relationship-dependent. Adoption of AI for direct counseling is minimal; the sector remains largely human-centric and skeptical of algorithmic substitution for therapeutic work.
Sector adoption velocityclaude-sonnet-52/5Social work and behavioral health sectors are slow adopters of AI for direct clinical/relational tasks, with pilots mostly limited to documentation or triage support.
Augmentation potentialclaude-haiku-4-5-202510013/5AI tools could assist social workers by drafting family education materials, organizing session notes, or providing evidence-based resource suggestions, but the core counseling and relational work must remain human-driven. Limited augmentation potential exists on the periphery.
Augmentation potentialclaude-sonnet-53/5AI can help social workers prepare psychoeducational materials, draft communication guides, or summarize case notes to inform family conversations, offering moderate assistance.
Task automatabilityclaude-haiku-4-5-202510011/5Counseling family members requires genuine empathy, understanding of complex interpersonal dynamics, and adaptive response to emotional nuance. Current AI cannot replicate the relational trust-building and contextual judgment needed to effectively support families, even with significant setup.
Task automatabilityclaude-sonnet-51/5Counseling family members involves nuanced emotional support, trust-building, and clinical judgment that current AI cannot perform end-to-end at equal quality with major time savings.
Adoption barriersclaude-haiku-4-5-202510015/5Mental health counseling is legally regulated; only licensed clinical social workers or equivalent credentials can provide therapeutic counseling in most jurisdictions. Liability and duty-of-care requirements create hard barriers that prevent full automation regardless of technical capability.
Adoption barriersclaude-sonnet-54/5Licensure, clinical liability, confidentiality (HIPAA), and the inherently relational/human-contact nature of family counseling create strong barriers to automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Even if narrow AI tools could assist, the oversight cost from a qualified social worker reviewing interactions would approach or exceed the full labor cost of direct human counseling, particularly given liability and error-cost asymmetry in therapeutic contexts.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute delivering this counseling output, so cost comparison favors the human by default.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs family counseling or therapeutic support at scale. While chatbots exist, they lack the clinical expertise, liability insurance, and demonstrated safety record required for real therapeutic interventions in mental health and substance abuse contexts.
Technical feasibility todayclaude-sonnet-51/5No deployed product independently counsels family members of mental health/substance abuse clients in production; this remains outside current AI product scope.

Collaborate with counselors, physicians, or nurses to plan or coordinate treatment, drawing on social work experience and patient needs.

1

CI 0–3 · exposure 0 · augmentation 50 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Healthcare sectors are cautious adopters of AI in clinical decision-making, especially for tasks requiring professional licensure and accountability. Pilots exist but displacement in actual treatment coordination is minimal.
Sector adoption velocityclaude-sonnet-52/5Healthcare and behavioral health are historically slower AI adopters for core clinical decision-making tasks, though administrative support tools are spreading.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist by retrieving patient history, suggesting evidence-based treatment options, or drafting coordination notes, but the core task of professional collaboration and clinical judgment remains human-centered.
Augmentation potentialclaude-sonnet-53/5AI can assist by summarizing patient records, flagging relevant history, or drafting care plan notes, but the collaborative decision-making itself is only lightly augmented.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires genuine human judgment integrating clinical expertise, patient-specific context, and nuanced professional coordination. AI cannot meaningfully participate in the collaborative treatment-planning conversation itself, which depends on real-time clinical assessment and interpersonal trust.
Task automatabilityclaude-sonnet-51/5This requires interpersonal judgment, real-time clinical collaboration, and integration of nuanced patient context that current AI cannot autonomously perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Licensed social workers, physicians, and nurses are legally required to participate in treatment planning and sign off on clinical decisions. Healthcare liability, regulatory oversight (HIPAA, state licensure), and the mandatory human accountability make automation legally and ethically infeasible.
Adoption barriersclaude-sonnet-55/5Licensed social workers, physicians, and nurses must legally participate in treatment planning and coordination; this is a regulated, credentialed clinical function with liability implications.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of AI oversight and error-checking for treatment coordination (given high error costs and need for human sign-off) would exceed the loaded wage of social workers who perform this task directly.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the human role in this task, so there is no viable cost comparison—human labor remains necessary.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs inter-professional treatment planning coordination. AI can draft notes or retrieve information, but cannot replace the actual collaborative clinical decision-making among licensed professionals.
Technical feasibility todayclaude-sonnet-51/5No deployed product coordinates multidisciplinary treatment planning with clinical judgment; existing tools only support documentation or scheduling, not the substantive collaboration itself.

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.