AI from slide to working software. In weeks, at a fixed price, with one accountable lead.
We find the AI use cases in your business that actually pay, prove the strongest one with working software inside three weeks, then ship it to production with a small senior team. Fixed price, agreed before anything starts.
Getting AI to change how the work actually happens is the hard part. That is where the budget goes quiet, the pilot stops being mentioned in the board pack, and the licences keep renewing anyway.
One bet at a time, chosen by guesswork
Most companies pick a single use case in a workshop, then spend two quarters finding out whether it was the right one. Too narrow to matter, too slow to learn from, too expensive to be wrong about.
Vendors sell tools. Agencies sell hours.
A vendor’s job ends when the licence is signed. An agency’s job ends when the timesheet is approved. Neither one is measured on whether the work actually changed.
The proof arrives after the money
You approve a business case built entirely on assumptions, then wait months to find out which of them held. By then the budget is committed and nobody wants to be the person who stops it.
None of this is expensive because AI is expensive. It is expensive because the decision gets made before the evidence exists. So we invert the order: evidence first, commitment second. That is the whole reason the Scan is short, fixed-price, and costs a fraction of what it stops you spending.
How to work with us
Three engagements, in order.
Start with the Scan. It is small and low-risk, and it de-risks the Build. The Build leads to Run. You can stop after any one of them — and you should, unless each step has earned the next.
Start here
01
AI Opportunity Scan
We find the use cases worth building, ranked by return and difficulty — and prove the top one with working software before you commit to it.
Interviews with the leaders who own the processes
A ranked agents catalog: every viable use case across the functions we cover, with ROI, feasibility and a build estimate
Four steps with dates against them. You see working software in week three of a relationship, not month six.
1Week 1
Discover
We sit with your leadership and the people who do the work, and map where time and money actually leave the business. You get the ranked roadmap in the first week, not at the end.
2Weeks 2–3
Prototype
We build a rough working version of the strongest candidate. Not a mockup — software you can click, shaped around your data — approved samples or synthetic data that mirrors it, never production access — doing the real task. This is what you use to decide.
3Weeks 4–13
Ship
Our delivery team takes the proven candidate to production: deployed on your infrastructure, monitored, documented, handed to your engineers. Fixed price, agreed before we start.
4Monthly
Run
We stay accountable for what is live and ship the next use case off the catalog — or we hand you the keys and step back. Both are good outcomes; only one of them is billable.
Week one starts with a conversation. Thirty minutes tells us both whether there is a week one.
None of these are positioning. Each one is the reason a specific part of the engagement is faster, cheaper, or less risky than the alternative you are comparing us against.
01
Evidence before commitment
Discovery ends with a working prototype, not a slide deck. You decide whether to fund the Build after you have used the thing — not on the strength of our estimate.
02
AI-native delivery
AI agents write the first draft of the requirements, the code and the tests. Senior engineers judge, harden and ship. That is why a small senior team moves faster than a traditional team of ten, and why we can quote a fixed price without padding it against our own uncertainty.
03
One accountable lead
One senior lead sits with your leadership, scopes the work, and answers for delivery. No account manager, no vendor-management overhead, and nobody handing you to a stranger after the sale.
04
Own your AI
Where it makes sense, we replace per-seat SaaS spend with an application you own. Your running cost becomes tokens and hosting — not a subscription that grows with headcount and reprices at renewal.
Why trust this
You should not have to take our word for it. So the engagement is built so you do not have to.
Every firm on your shortlist will tell you they are senior, fast and different. None of them can prove it before you pay, and neither can we. So rather than ask for the benefit of the doubt, we made the first engagement the cheapest possible way to find out.
The Scan is three weeks at a fixed fee. It ends with software you have actually used, a catalog you keep, and a priced Build scope you are free to hand to somebody else. If we are not good, you will know inside a month — for a fraction of what discovering it during a Build costs.
That is the proposition. Not credentials — structure.
Fixed price.If the work is harder than we estimated, that is our loss — not a change request.
You own it from day one.Code, infrastructure definitions and documentation live in your accounts, not ours.
No obligation after the Scan.The catalog, the process map and the prototype are yours either way.
Nothing locked in.Run is rolling monthly, cancelled with 30 days notice. There is no annual commitment anywhere.
What we will show you on the call
We have no logos to show you yet. What we can do is walk you through the artefacts themselves:
The scoring model behind the agents catalog — how candidates get ranked on value, feasibility and data readiness, and how a build estimate is derived from that.
A prototype at the fidelity you would get in week three, so “rough working software” is not left to your imagination.
The scope document and acceptance criteria a fixed price is written against.
Thirty minutes, under NDA if you prefer.
Questions we get asked
The things you would ask in the first ten minutes.
What does “fixed price” actually cover?
A written scope agreed before work starts, with named deliverables and acceptance criteria. If the scope holds, the price holds — including the weeks where the work turns out harder than we estimated. That risk is ours, and it is priced in. If you want something outside the scope, we agree it as a change before we build it. You will never see an invoice you have not already approved.
Who actually does the work?
A senior lead runs the discovery, the scoping and the leadership conversations, and stays on the engagement from first call to handover — that part is never delegated. For a Build we add senior engineers we have worked with before and hand-pick per engagement. No juniors learning on your budget, no bench to keep busy, and nobody standing between you and the people writing the code.
What if the prototype does not work?
Then you found that out in three weeks for the price of the Scan, instead of finding it out at the end of a six-figure build. That is the Scan doing its job. A catalog typically holds eight to fifteen ranked candidates, so a weak leader almost never means a weak list — we move down it. And if the honest answer is that AI is the wrong instrument for your problem, we write that down and say so.
How do you handle our data and security?
Everything runs inside your environment and under your policies by default: your cloud, your identity provider, your retention rules. We use commercial model endpoints with training on your data and code switched off, and we sign your NDA, DPA and security addendum rather than asking you to accept ours. We build to financial-services expectations by default — least privilege, audit trails, and no data leaving your boundary — so your review finds what it expects to find rather than a list of exceptions.
Which industries do you work in?
We focus on financial services — wealth and asset management, insurance, fintech and the professional services around them — because the constraints there are the hardest and the processes are the most expensive when they go wrong. The method itself is industry-agnostic. What matters more than your sector is that you have roughly 50 to 500 people, a process that visibly costs you money, and a leadership team willing to give us a few hours in week one.
How does an engagement start?
With a 30-minute call. You describe one process that costs you money; we tell you whether AI is the right instrument for it and what a Scan would look like. If it fits, you get a one-page scope with a fixed price and a start date. Most engagements begin within two to three weeks of that call.
Next step
Thirty minutes is enough to know if this fits.
Bring one process that costs you real money. We will tell you on the call whether AI is the right instrument for it, what it would take to build, and roughly what it would cost. If the answer is no, we will say so — that is a cheaper outcome for both sides than a proposal neither of us believes.