← Practice Learning Hub
Free Learning Resource

AI in Social Work: Opportunity, Risk & Professional Judgement

A practice-led learning pathway for thinking critically about where AI can help, where it can create new risk, and how to keep people, ethics and professional judgement at the centre.

30 / 60 / 90 min + videoTeam learningAI literacy & ethicsIndividual CPD
Important: This is an educational resource, not approval to use a particular AI tool. Use fictional or fully de-identified examples in discussion. Never put identifiable case information into a public or unapproved AI service. Follow your employer’s current AI, information governance, data protection, recording and professional practice requirements.

Learning Outcomes

  • Identify common ways AI is already being used or proposed in social work practice
  • Explore potential benefits such as reducing administrative burden and improving accessibility without assuming that efficiency equals better practice
  • Recognise risks including confidentiality, inaccurate or fabricated output, bias and discrimination, loss of nuance, over-reliance and reduced transparency
  • Apply professional judgement and meaningful human oversight to AI-assisted work
  • Use the PAUSE framework to decide what questions to ask before using or relying on AI in practice
  • Identify one action for local policy, team learning or individual CPD

Step 1: Watch — AI in Social Work: Opportunity or Massive Risk?

Kayleigh explores AI entering social work through recording, case notes, assessment and other forms of practice support. Watch for both sides of the tension: what could genuinely improve practice, and what could become less visible when an AI output sounds polished and confident.

Learning videoAI in Social Work: Opportunity or Massive Risk?Play in privacy-enhanced YouTube mode

While you watch, notice:

  • Which examples feel like genuine augmentation of social work, rather than replacement of professional judgement?
  • Where could speed improve while voice, trust, context or quality becomes weaker?
  • What assumptions might be hidden behind a fluent AI-generated record, summary or recommendation?
  • Which decisions or tasks do you think must remain distinctly and visibly human?

Reflective Prompts

Use these after watching. Keep examples fictional or fully de-identified and distinguish what a system can do from what your employer has approved it to do.

Which uses of AI in the video feel like genuine augmentation rather than replacement of social work?
Where might efficiency improve but practice quality, the person’s voice or trust reduce?
What would you not be comfortable delegating to AI, even if the tool was employer-approved?
How might a polished AI output make error, bias or missing information harder to notice?
What needs to remain visible and auditable when AI has contributed to a record, assessment, summary or recommendation?
What should a person using services reasonably understand about AI involvement in work that affects them?

PAUSE Before AI

A five-question practice check for any AI-assisted task. It is intentionally simple enough to use in a team conversation, while pointing back to local governance and professional responsibility.

P — Purpose. What problem are we trying to solve? Is AI necessary and proportionate? What should improve for the person or the quality of practice — not only speed?
A — Approved & accountable. Is this tool approved for this use? What policy applies? Who remains responsible for the final record, assessment, recommendation or decision?
U — Understand the data. What information is going in? Is it personal or special-category data? Where is it processed, retained or reused? Are we using only what is necessary?
S — Scrutinise the output. What could be inaccurate, invented, biased, missing or over-confident? Can important statements be traced back to evidence and checked?
E — Explain & keep the person central. What does the person need to know? Does the process preserve participation, voice, privacy, relationship and meaningful human judgement?

Fictional Practice Scenarios

The scenarios are deliberately incomplete. The aim is not to find one correct answer, but to identify what is missing, what the risks are, what safeguards are needed and what remains the practitioner’s responsibility.

Scenario 1 — The ambient note-taker

A local authority has approved an AI tool that can listen during visits and draft a case note. A social worker explains it at the start of a home visit. The person agrees, but becomes visibly quieter once recording begins. Halfway through, an adult family member enters and a sensitive disclosure follows. The AI produces a fluent summary that sounds accurate at first glance.

What needs to be considered about transparency, participation and the change in context?
What would you check before accepting the generated note?
Could the presence of the tool change the relationship or what a person chooses to say?

Scenario 2 — The polished assessment

A practitioner uses an employer-approved generative AI assistant to help structure an assessment from de-identified notes. The draft is clear and saves time, but it introduces a confident explanation for missed appointments, softens some of the person’s own words and gives less space to strengths and cultural context than expected.

Which parts of the output require evidence checking?
How might bias appear even when the language sounds professional?
Who authored the final assessment, and what does that mean for accountability?

Scenario 3 — AI-supported prioritisation

A service is considering a system that summarises referrals and flags cases it predicts may need urgent attention. Practitioners are told a social worker will still make the final decision, but the score and summary will appear prominently at the top of each referral.

How might the recommendation influence human judgement even if it is not formally “the decision”?
What groups could be disadvantaged if historical data contains unequal patterns?
What would meaningful human oversight and a challenge route look like?

Choose Your Session Length

These facilitated timings assume participants have watched the video beforehand. If you watch together, add the video runtime or shorten the discussion activities to fit your slot.

30-Minute Reflective Discussion

Best for: team meetings, quick CPD slots

0-5 min
Open: what AI is already present in our work, and what feels helpful, uncertain or uncomfortable?
5-15 min
PAUSE: introduce the five questions and apply them in pairs to one AI use participants have encountered or heard about.
15-25 min
Scenario: use Scenario 1 or 2. Identify one potential benefit, one risk and one safeguard.
25-30 min
Action: each person completes: “One thing I will check before I use or rely on AI is…”

60-Minute Team Session

Best for: team development, ASYE groups, practice education

0-8 min
Temperature check: where is AI already touching our workflow? What should remain distinctly human?
8-20 min
PAUSE: work through each question and connect it to local policy or current practice.
20-40 min
Scenario groups: different groups take different scenarios and identify benefits, harms, missing information, safeguards and accountability.
40-52 min
Whole-group debate: “Efficiency is not the same as quality.” What would good AI-assisted social work look like? What would poor practice look like even if the output sounds polished?
52-60 min
Practice commitment & CPD: agree one team action and record one individual learning point.

90-Minute Development Session

Best for: learning events, workforce development and facilitated workshops

0-10 min
Open: set ground rules and map current or proposed AI use across the team.
10-25 min
PAUSE deep dive: distinguish technical capability from professionally acceptable use.
25-55 min
Scenario carousel: rotate through all three scenarios, adding risks, safeguards and questions for the person, practitioner, manager and organisation.
55-72 min
Local workflow audit: choose one ordinary workflow, not a live case. Where could AI assist? What new risk could it introduce? What approval, training or transparency is needed?
72-83 min
Green / amber / red: agree potentially appropriate uses, uses needing extra checks, and uses that should not happen without formal approval.
83-90 min
Close: capture one individual CPD action and one organisational question to take forward.

Download Resources

Everything needed to turn the video into a structured learning session. All scenarios are fictional and the materials are free to use for learning and reflection.

📖

Facilitator Guide

PAUSE framework, 30/60/90-minute plans, three scenarios, facilitation cues and current source material

Download PDF
📝

Participant Worksheet

Pre/post reflection, PAUSE questions, green/amber/red boundaries and scenario worksheets

Download PDF
📑

CPD Reflection Template

Typeable Word template connecting AI literacy and professional judgement to CPD

Download Word

Standards & Guidance This Resource Supports

Mapping is indicative. AI policy and guidance are evolving quickly, so use the links below alongside your employer’s current arrangements.

Social Work England professional standards
Particularly Standards 2 (dignity, privacy and confidentiality), 3 (legal/ethical frameworks, evidence, bias, professional judgement and ICT), 4 (CPD), 5 (professional integrity) and 6 (ethical practice).
Current social work AI evidence
Social Work England’s AI research and BASW’s GenAI statement provide profession-specific context on current use, opportunity, risk, governance and keeping people and professional judgement central.
Data protection & governance
ICO guidance on AI and data protection, UK data protection law, information governance and your employer’s approved-tool and recording arrangements remain central.

CPD Reflection Prompts

Use these to turn the session into recordable professional development.

What did I learn about how AI is already being used or proposed in social work?
Which opportunity for AI seems most useful in my role, and what evidence would I need that it improves practice rather than only speed?
Which risk or ethical tension most challenged or developed my thinking?
How will I use the PAUSE questions before I use or rely on AI?
How will I make sure professional judgement, the person’s voice and meaningful human oversight remain central?
What will I do differently, check with my organisation or learn more about after this session?

Using This With Your Team

This pathway works particularly well where AI is beginning to enter practice but teams are at different stages of confidence. It is designed to make uncertainty discussable without turning the session into either an AI sales pitch or a blanket warning against technology.

Keep the conversation grounded in ordinary workflows, approved tools, professional responsibility and the person’s experience. Where a live concern emerges about confidentiality, data protection, policy or a particular AI deployment, move it into the appropriate organisational route rather than trying to resolve it through the learning exercise.

Using this resource?

If your team, authority or university uses this pathway, Kayleigh would love to hear what was useful and what the next AI learning question should be.

Share Your Feedback

Prefer Kayleigh to deliver it live?

Kayleigh speaks at a select number of events each year and is available on request for keynotes, workshops, webinars and tailored learning sessions.

View speaking & workshop options

Educational resource only — follow current employer policy, data protection requirements and authoritative professional guidance. Scheduled for review February 2027.