Beyond prompting: a lawyer's checklist for steering AI agents through real work
jkubicki · x · 2026-11-05
The author, who teaches law students and lawyers to use AI, argues the skill that matters is "steering" — managing the AI's work process like a subordinate — not prompting. His checklist:
Before starting
- Frame the problem: what question, for whom, why
- Set the method: issue-by-issue, element-by-element, or comparison
- Bound the search: which sources are allowed, off limits, how far back
- Assign tools to steps
- Sequence the work and set checkpoints
- Define done as a test the AI can grade itself against
While it runs
- Inspect intermediate steps, not just the final draft — that's where errors hide
- Redirect when it drifts instead of blindly regenerating
Context: the GDPx benchmark showed frontier models score near 0% on real expert legal work, so steering skills matter more than prompting.
More from coding & agent
- Dots Team Recruits to Build Out Persistent Agents — sherwinwu · 2026-11-05
- Managing agent workstreams as infrastructure-as-code in production — JoshPurtell · 2026-11-05
- AI Coding Tool Conductor's Team Joins Meta to Build the Next Interface — ycombinator · 2026-11-05
- 'You own every line you submit': the coding policy that survives AI — esrtweet · 2026-11-05
- Notion AI agents gain Google Drive access: search files, edit Drive, trigger workflows — technextpreneur · 2026-11-05
- Dev builds workflow that auto-files a GitHub issue whenever he's the blocker — SubtleGradient · 2026-11-05