AI prompting and bias training for legal and governance teams
Your people have the AI. This is the training that makes them safe and effective with it: how to prompt well, verify efficiently, and recognise the moments when human bias and machine bias combine to produce answers that feel checked but aren't. Interactive, practical, and model-agnostic.
The skill gap isn't prompting. It's judgement
Most AI training teaches syntax — magic phrases for better outputs. Useful, but it misses where legal and governance work actually goes wrong: users write leading questions and the AI politely agrees; fluent answers earn unearned trust; verification is skipped when the checking is hard. Researchers now call this compound human–AI bias, and it's the failure mode our training is built around. Participants learn to prompt for accuracy rather than agreement, to verify against the source quickly instead of exhaustively, and to spot the interaction patterns that mislead — skills that transfer across every model and tool, which is why our sessions are deliberately vendor-agnostic.
Session formats
Interactive workshops (half-day or full-day) using your document types and realistic scenarios; executive briefings for boards and leadership on governing AI over the organisation's knowledge; and train-the-trainer programs for legal ops and knowledge teams embedding capability internally. All sessions are built by the team the FT has recognised for AI work every year since 2022 — and informed by what we see in live user testing of people using AI on legal documents, which is where the real failure patterns come from.
Who it's for
In-house legal, risk, compliance and governance teams; law firms rolling out AI platforms; and government agencies deploying assistants over policy and legislation. Our programs were specifically developed for risk-averse, highly analytical professionals — we've been training lawyers in new ways of working since our legal design training began in 2018.
FAQ
Is the training specific to one AI tool?
No — deliberately. We train judgement and technique that transfer across Microsoft Copilot, legal AI platforms and general assistants. Tool-specific tips date in months; prompting and verification discipline doesn't.
What is compound human–AI bias?
The interaction of human cognitive biases (leading questions, over-trust of fluency, confirmation-seeking) with AI system biases — each amplifying the other to produce confident, wrong outcomes. Our training teaches participants to recognise and break the loop.
Can sessions use our own documents and scenarios?
Yes, and they should — training lands hardest on your real contract types, policies and typical queries. We build exercises from your material under confidentiality.
Do you offer CPD/CLE accreditation?
Sessions are designed to qualify for CPD in most Australian jurisdictions as substantive professional skills training. Ask us about your jurisdiction's requirements when booking.