Training for Professionals
The Programme
Foundations first. Live demos against your work second. The gap is where learning happens.
2 hours
2 hours
Prompt Engineering
The training covers prompt engineering as a discipline, not a trick. Four patterns lawyers use every day.
Live Demos
Three example demos other firms have used. Session 2 runs live against your firm’s own material — pick from these, or describe a workflow you’d like to see demoed against.
Bring your own. Tell us what your team wrestles with and we’ll demo against it in Session 2. The three above are starting points, not the menu.
Pricing
Each session is two hours, delivered remotely. Most firms book the two-session programme. Wrap-around services scoped separately.
Standard team
Up to 20 attendees
£3,500
both sessions
Larger team
20 to 50 attendees
£5,500
both sessions
Bespoke
50+ attendees
Tailored
scoped on enquiry
What comes after training?
See the full tier ladder →Session 1 / Foundations
How the models work, where they fail, and how to set up Claude for legal practice. Regulatory and court guidance included.
Session 2 / Going Deeper
Advanced settings, skills, and three live demos against material the firm brings. Answers the questions the practice gap surfaced.
Pattern 01 / Foundation
A working prompt is rarely a single sentence. It carries five components: Role, Context, Task, Format, Constraints. Skip any one of them and the output drifts.
Same task, written twice
Before training
What comes back
A generic letter with no understanding of position, jurisdiction, client, law, or firm tone. Plausible. Not usable.
After training
What comes back
A jurisdictionally accurate, properly headed, strategically calibrated first draft. Ready to review and send.
Pattern 02 / House style
Describing your firm’s tone in words rarely lands. Pasting two or three real examples of the output you want lands every time. The model matches what it sees over what it’s told.
Same task, two ways
Description only
What comes back
Adjective soup. The model interprets “warm but professional” one way. Your firm interprets it another. Mismatch.
With examples
What comes back
A new update that reads as if the same hand wrote it. Sentence rhythm, vocabulary, structure all match.
Pattern 03 / Iteration
The first response is almost never the final one. Specific feedback closes the gap faster than a new prompt. The model carries the previous context forward.
A typical refinement chain
Three turns. Each one a specific change. Total time: under five minutes. The judgment lives in your edits, not in the typing.
Pattern 04 / Hallucination guard
The largest single risk in legal AI work is confident wrongness — invented cases, plausible but false citations, inferences treated as fact. Explicit constraints reduce it dramatically. They flag the model into a more conservative mode.
Constraint examples worth using
The model will refuse, flag, or qualify rather than guess. Output becomes more cautious — exactly the disposition required for regulated work.
Session 2 demo / Live
A new matter lands. The clock is running. Here is how to get to a structured first view, fast, without cutting corners on review.
Session 2 demo / Live
Staged drafting from a firm template with live variation instructions. Shows the iterative workflow, from template to matter-specific draft without starting from scratch.
Session 2 demo / Live
Financial disclosure is often where matters drag. This demo shows how to combine tool output with Claude analysis for a seamless pipeline from disclosure to findings.