Proof of concept
A validated proof of concept in ninety hours
An Asian bank wanted to know whether an AI copilot would help its relationship managers. The first iteration took nineteen hours; the whole validated proof of concept took ninety and scored four and a half out of five on acceptance.
- First iteration in front of users
- 19 hoursFirst iteration in front of users
- Total for the validated proof of concept
- 90 hoursTotal for the validated proof of concept
- Client acceptance score
- 4.5 / 5Client acceptance score
The client
An investment-banking relationship-manager function inside an Asia-Pacific bank, reached through a specialist AI partner. Data spread across a core banking platform, a document repository, spreadsheets and a market terminal.
The engagement
An eight-week proof of concept with a formal acceptance score at the end — deliberately small enough that the client could say no cheaply.
A working demo of this build exists.
It is not public yet — ask for access, and where the NDA allows I will send a link or walk you through it on a call.
The problem
Banks routinely spend six figures discovering whether a copilot helps a role, and the discovery itself produces a document rather than a working answer. The risk is not that the technology fails — it is that a large validation budget commits everyone emotionally before anyone has seen the thing work, which makes an honest no almost impossible.
What I did
I made saying no cheap. The first iteration went in front of users after nineteen hours, at interface fidelity, because the fastest way to learn whether a copilot fits a workflow is to let the person whose workflow it is press the buttons. Scope was held to four capabilities aimed at one role rather than a platform aimed at a department, and both an experienced and a junior relationship manager were treated as distinct personas, since the value of a copilot differs sharply between someone who knows the answer and someone who does not. The whole thing closed with a formal acceptance assessment, so the outcome was a score the client could act on rather than a positive impression that evaporates by the next budget cycle.
What was built
Four narrow capabilities aimed at a single role — a market copilot answering questions over investment-office documents, a proposal personaliser, a client context builder tracking preferences and insights, and an opportunity radar alerting on market movements — prototyped at interface level first, with both experienced and junior relationship managers as explicit personas.
On the table at the end
- Interactive proof of concept covering four capabilities
- Interface prototypes validated with two user personas
- Formal acceptance assessment
What it changed
Answered a capability question for the cost of an opinion. Ninety hours produced a formally accepted proof of concept instead of a six-figure pilot, and gave the bank an evidenced basis for deciding whether to go further.
How it ran
- 01
Nineteen hours to first contact
Interface-fidelity iteration in front of real users before any backend investment, because workflow fit is discovered by use, not by interview.
- 02
Four capabilities, one role
Copilot, proposal personalisation, client context and opportunity alerts — narrow enough to prove, wide enough to matter.
- 03
Two personas, deliberately
Experienced and junior relationship managers evaluated separately, because a copilot's value inverts between them.
- 04
Map the real data sources
Core banking platform, document repository, spreadsheets and market terminal identified up front, so the proof of concept knew what it was pretending about.
- 05
Close with a score
A formal acceptance assessment rather than a warm meeting, so the go or no-go rests on something written down.
Other work
All case studies →- Corporate banking
Winning back the last call after two meetings had failed
A long-standing account was one bad meeting from closing. I rebuilt the pitch around the client's own operating numbers and a working prototype, and cut every claim we could not source.
- Regional banking
One prototype, two markets, no budget in the room
A meeting with no project, no budget and no agenda. I brought a day-in-the-life prototype in the client's own language rather than a capability deck, then shipped their feedback as working features.
- Aviation
Twenty days of discovery that priced a year of build
An airport group needed to replace a security-credentialling system under a sovereign compliance regime. Twenty person-days of discovery produced five approved deliverables and an effort envelope the fixed-price contract could stand on.
Something similar on your plate?
Thirty minutes, no deck. I will tell you whether it is worth doing at all.