Taught from practice
Trainers who use these systems on real legal work daily — the failure modes shown are the ones actually encountered, not slideware hypotheticals.
AI training and supervised rollout from Aun & Co.: hands-on adoption for working teams, taught by a practice that governs its own AI use daily.
Tools and policies fail at the same point: the person at the keyboard on a Tuesday afternoon. Training and rollout is the phase where governance becomes habit — the team taught not just the rules but the craft: what these systems do well, where they fail, how to review an output that looks right, and what never to paste. The firm teaches from its own floor: the workflows, mistakes and review habits of a legal practice that adopted AI under human-approval gates and kept them.
Training is built on the team's real documents and real tasks — generic demos train nobody. Each role learns its own permitted workflows, then the review craft: verification habits for the failure modes that matter in professional work, where a fluent wrong answer is the danger. Go-live is phased — sandbox, supervised live, checkpointed independence — and the rollout ends with a review that feeds reality back into the policy, because the first month always teaches something the draft missed.
Trainers who use these systems on real legal work daily — the failure modes shown are the ones actually encountered, not slideware hypotheticals.
The core craft of professional AI use — verifying outputs that look correct — taught deliberately, with exercises on your own document types.
Champions, escalation paths and a feedback loop installed, so the capability keeps improving after the engagement ends.
A typical engagement: a team with approved tools and a policy but thin adoption. Role-based sessions on the team's own documents, two weeks of sandboxed practice, supervised live use with checkpoints — and by the post-rollout review, usage has moved from two enthusiasts to a governed team habit.
Described in abbreviated, anonymised form to preserve client confidentiality.
Three layers: the rules — tools, inputs, gates from your policy; the craft — prompting, and above all reviewing outputs that look right; and the workflows — each role's actual permitted uses practised on its actual document types. The middle layer is the one generic training skips.
For a small-to-mid-size team: training across a week or two, a sandbox fortnight, then supervised live use with checkpoints over a month. A quarter end to end, roughly — governed habit takes repetitions, not memos.
Against the value case and the incident log together: usage rates by workflow, time actually saved, review-gate compliance, and zero uncontained input incidents. The post-rollout review reads all four and adjusts the policy to what the first live month revealed.