The Legal Practice Council hasn't written formal AI guidance yet, but the existing professional conduct rules apply directly. Done carefully, AI saves attorneys 10–15 hours/week. Done sloppily, it creates professional-liability exposure. Here is the line.
Safe AI uses for attorneys
- Client intake chatbots for initial enquiries and conflict checks
- First-draft contracts and NDAs for attorney review
- Summarising lengthy depositions or discovery documents
- Automating billing narratives and time entries
- Client status communications ("Your matter is in X stage")
Unsafe AI uses
- Giving legal advice directly to a client through AI
- Filing AI-generated court documents without line-by-line review
- Using public AI tools for privileged client information
- Relying on AI-cited case law without verification (several cases have already been dismissed for fictional citations)
The confidentiality stack
For attorneys, we only deploy AI with: zero-retention API settings (the AI provider stores nothing after the conversation), EU data residency or SA-based processing, and explicit matter-specific separation so one client's data can't leak into another client's context.
Time recovered in practice
A solo attorney running AI-assisted intake and document workflows typically recovers 12–18 billable hours/month. For a 5-attorney firm, that scales to 60–90 hours — which is usually one additional effective fee-earner in capacity.
The attorneys who get AI right in 2026 will be the ones who treat it like a paralegal who needs close supervision — useful, but never final.