AI transformation
The AI roadmap your board can fund.
Six weeks, kickoff to board readout. A paid engagement that ends in a costed plan your own tech, ops and finance teams have signed. No build inside it.
The stall
The budget grew. The return didn't follow.
15%
of organizations piloting AI report any revenue growth from it.
85%
see no measurable revenue impact from AI at all.
Grant Thornton · 2026 AI Impact Survey · n=950
Everyone has Copilot now.
And no one can point to the P&L line it moved.
We've mapped hundreds of workflows. Where do we even start?
And no one has put an ROI on any of them.
The demo was brilliant. But will it work on our legacy systems?
And no one can answer that yet.
The pilot scaled. So did the bill.
And the P&L case no longer closes.
Underneath all four: where does AI actually pay, and will it work here?
The proof
A $2.6B US fintech asked us those questions.
Core systems 20+ years old · a product estate assembled by acquisition · regulated on every side: SOX, PCI, GDPR.
01
Licenses were already bought and pilots had run, and none of it had a number against it.
02
Every team was spending separately, with no shared plan and no single number to show for it.
03
The board wanted a plan it could approve, not another round of pilots.
6 wks
Kickoff to board readout
13
Processes mapped across four departments
~$8-16M
Annual value identified, against ~$2M to build it
These guys are from the future - they're living it. Not the same old consultants handing you a deck you'll never open again.
AI transformation leader · a $2.6B US fintech
We build what we scope. One of ours: AI email triage, 76% faster case resolution (8.8 → 2.1 days).
The method
Read · Rank · Fund.
Three steps, about six weeks, one costed roadmap. Each step hands you something you can act on before the next one starts.
-
01
Read
We read your code, docs, tickets and logs.
You get a measured baseline.
-
02
Rank
We rank your workflows for highest return and lowest risk.
You get an ordered short list.
-
03
Fund
We price the short list, after sign-off by your own tech, ops and finance teams.
You get a costed, board-ready roadmap.
Read
We measure your baseline, not just ask about it.
Most assessments stop at a questionnaire. We go on to your code, your tickets and your logs, and measure what they say.
-
01 Your people
Ask the people
We interview heads, owners and frontline staff, and find the champions.
-
02 Your documents
Read the org's memory
We ingest SOPs, runbooks, tickets, wikis and post-mortems.
-
03 The real work
Watch the real work
We watch the work over the shoulder, workarounds included.
-
04 Your systems
Agents decode your systems
Our agents read your code, data and logs for real cycle times.
RankWorked example
Four tests decide the order - not the loudest room.
The order the four tests produce on their own, and where a budget cycle draws its line. The workflows and scores are an illustration, not a client's roadmap.
-
01 Data
Is it there, and can our agents reach it?
-
02 Adoption
Is there an owner who wants this to work?
-
03 Metric clarity
Is there one number, already tracked, to prove it moved?
-
04 Impact
Is the prize big enough to move your operating number?
The order we'd build in Score / 16
- 1 Payment exception triage 16
- 2 Customer onboarding 15
- 3 Invoice matching (AP) 14
- This cycle's budget stops here
- 4 Reconciliation 13
- 5 Dispute handling 11
- Not worth building
- 6 Contract review 9
- 7 Vendor KYC refresh 8
Fund
Every line arrives signed by your own teams.
Each roadmap item ships as a full design report, and the people who signed it work for you, not for us.
-
01 Your architects · it fits the stack
The solution diagram is drawn on your actual legacy systems, and approved before it's priced.
-
02 Your operations team · people will use it
The workflow re-designed with the people who run it, adoption planned in.
-
03 Your security and risk team · it clears review
Data paths, access and retention reviewed against your own controls before a line is built.
-
04 Your FP&A · the numbers hold
The projected ROI approved by finance, with our opex projections alongside.
Design report · v0One roadmap line
- The problem, and how the work runs today
- The solution, drawn on your actual stack
- Data quality, graded, and how agents get to it
- Build plan, team, timeline
- Running cost at volume, three scenarios
- Hours given back to the team, methodology shown
- What must be true, and what could go wrong
All seven sections, per line item, in writing. The roadmap is build-agnostic: it holds its value whoever builds it.
Bring your two most contested AI bets.
One working session, and we'll show you how they score. If neither is worth building, we'll tell you that instead.