Six weeks inside one practice area, finding where the team and its clients feel the pain and what would fix it. It ends in a shortlist of priorities, tested with the people who do the work.
6 weeks1 practice area1 validated shortlist
Weeks 1-2 · Product thinking, then the map
Every idea in the practice, from all levels
A week learning to think like a product team - problems before features, and what AI really does today - then a workshop mapping how work actually moves through the practice: where time leaks, where clients wait, where margin quietly goes.
Weeks 3-4 · Interviews
Pressure-tested by the people who know
Interviews with the people closest to the work to go deeper: your partners, associates and PSLs of the practice. We use this to shortlist the ones that are worth solving, and those without an obvious fit are parked for later.
Week 5 · Reality check
The strongest get circled
Does the data exist, would users switch, can it be built inside a quarter, will someone pay? Most ideas stop here, and that is the point.
the ones worth building
Week 6 · Prioritise, plan, present
A shortlist and a 90-day plan
A discovery report, an AI ideas pack and a 90-day plan the practice can act on. The sprint in full follows below.
0102030405
Scroll
The sprint · Six weeks
Six weeks to know what to build first, and why
Every practice group has a wall of AI ideas. What it lacks is a filter. You finish with the few worth backing, each written up with the problem, the likely saving and the rollout effort, and a 90-day plan to start the first pilot.
Phase 01 · Week 1
Learn to think like a product team
Before anyone suggests a tool, the group learns how product people think: start with the problem, name who has it, and agree what success looks like. Alongside it, an honest briefing on what AI does well today and what is still hype - so every idea that follows starts from reality.
Design thinking for lawyers: problems before features
How to frame a problem worth solving
What AI does well today, and what is still hype
The product lens · CorporateWeek 1
The request
“We need an AI toolwhich problem? for due diligence.”
Reframed as a product problem
WhoAssociates on M&A deals
The painTwo days triaging every data room before the real review starts
Done whenRed flags surfaced before kick-off
Real today
Pulling clauses outFirst draftsSummaries
Still hype
Replaces the associateNo checking needed
Purple · Discovery sprintProblems before features
think like a product person
Phase 02 · Week 2
Map the friction in real workflows
A workshop maps how work actually moves through the practice - where time leaks, where clients wait, where margin quietly goes. How-Might-We reframing turns complaints into well-formed product openings.
Live workflow mapping with the people who do the work
Friction reframed as How-Might-We questions
A defined set of opportunity areas out of the room
From a real workshop
From the board · FundsWeek 2
side lettersHow might we track every MFN obligation across side letters in minutes?
LPA mark-upsHow might we check a new LPA against house terms before the first turn?
investor DDQsHow might we answer a DDQ from past answers in a morning?
a chatbot for everything?
Phase 03 · Weeks 3-4
Pressure-test with the people who know
Deep-dive interviews with the partners, associates and PSLs of the practice - validating assumptions, surfacing what a workshop cannot, and hearing exactly where adoption would live or die.
Structured stakeholder interviews
Assumptions checked against daily reality
Hidden opportunities surfaced from the floor
“We still read every trade mark watch report line by line.
Senior associate · IP
“Renewal deadlines live in three systems and a spreadsheet.
PSL · IP
“Clients want one view of their whole portfolio, and nobody can give it to them.
Partner · IP
Phase 04 · Week 5
Reality-check what survives
Every surviving idea meets the same four questions: does the data exist, would users switch, can it be built inside a quarter, and will someone pay? Ideas that pass advance. Ideas that fail are stopped cheaply, on paper.
Feasibility, data and adoption checks
Market signal tested against our build experience
Weak ideas stopped before they cost anything
Does the data exist?
Would users switch?
Built in a quarter?
Will someone pay?
Lease abstracting
Know-how search
Report on title drafts
Planning search summaries
Rent review tracker
Break-date alerts
Completion bible builder
Advance
Does the data exist?
Would users switch?
Built in a quarter?
Will someone pay?
Lease abstractingKnow-how searchReport on title draftsPlanning search summariesRent review trackerBreak-date alertsCompletion bible builder
killed cheaply, on paperExample · Real estate
Phase 05 · Week 6
Prioritise, plan, present
The sprint ends in a leadership session: the shortlist, the reasons behind it, and a 90-day plan for the top pilots with clear go and no-go checkpoints. A path forward the leadership has already bought into.
Ranked shortlist with the scoring rationale
90-day quick-start plan with checkpoints
Presented to leadership for a decision
The shortlist · DisputesWeek 6
01Chronology builder8.9
02Disclosure summaries8.4
03Bundle pagination7.6
Purple · Discovery sprint90-day plan attached
What you leave with
A shortlist with its homework done
Discovery report
A plain-English account of the bottlenecks, the opportunities and the priorities, with success measures attached.
AI ideas pack
The strongest opportunities, one page each: the problem, how it works, the likely saving, the rollout effort.
90-day plan
A practical plan for the first pilots, with go and no-go checkpoints designed to build momentum fast.
Discovery report
AI ideas pack
90-day plan
built to start, not to file
From a real sprint
One firm stopped debating AI ideas and started building them
An earlier, firm-wide version of the sprint for a fast-growing UK specialist firm. Every idea was scored on value and feasibility, leadership made a clear call on direction, and the 90-day plan had the firm writing code within four weeks of the engagement ending. And because we build what gets chosen, every idea is scored against what building it actually takes.
