AI as a drafting aid, not a decision-maker
Social workers and case managers spend a large share of their week on documentation — case notes, assessments, referrals, and reports — rather than with the people they serve. AI can give some of that time back by structuring what you already know into clear, consistent prose. The crucial framing is that AI is a drafting and organising aid: it never observes a family, never weighs risk, and is never professionally accountable. You are. Used within that boundary, it reduces administrative drag without touching the judgement that defines the profession.
How AI helps with case notes
The most reliable use is turning rough, factual notes into structured records. You dictate or type what happened — who was present, what was said, what you observed — and ask the model to organise it under standard headings such as presenting issue, observations, actions taken, and next steps. The rule that protects you is simple: the model may only reformat facts you supply, never add detail. If you write “child appeared withdrawn,” the output should not embellish that into a diagnosis. Always read the draft against your raw notes before saving, and keep your own factual log as the source of truth.
A second strong use is risk assessment scaffolding. A model can produce a checklist of indicators relevant to a domain — neglect, domestic abuse, self-neglect in older adults — prompting you to consider each one. This is a memory aid against omission, not a scoring engine. The weighting, the context, and the safeguarding decision remain yours.
Data privacy, safeguarding, and disclosure
This is where most of the risk sits. Identifiable client information must never go into a public consumer chatbot, which may retain inputs and train on them — that is an unlawful disclosure and a safeguarding breach. Work with anonymised placeholders (“Client A,” “the older parent”) and only insert real identifiers locally, in your agency’s approved systems, after drafting. Better still, use only tools your employer has formally assessed for data residency, retention, and confidentiality under UK GDPR or your jurisdiction’s equivalent.
For referral letters, AI is genuinely useful: feed it the anonymised facts and the recipient (GP, housing team, CAMHS) and it will produce a clear, professional draft you then personalise and verify. Disclose AI assistance per your employer’s policy, keep the human-reviewed version as the official record, and never let generated text stand in as a verbatim account. The discipline is consistent throughout: AI accelerates the writing, you own the judgement, the facts, and the accountability.