Case · ABN AMRO
James ran the first ChatGPT pilot inside ABN AMRO. A walled-off, local model let staff find out for themselves what AI does for them. Earlier, at the same bank, he mapped the client screening process. Two stories, one lesson: understand first, scale after.
The story
A bank is one of the most tightly regulated places to work there is. Regulators look over your shoulder, client data is strictly protected and the reputation sits under a magnifying glass. At the same time, ChatGPT had brought a new generation of AI to the door. The question was never whether the bank should do something with it, only how that could be done safely.
A pilot with two clear boundaries, in place of a bank-wide rollout. First, a walled-off local model: no data went to an outside party. Second, only processes without direct client contact, so no client could be affected while the team learned.
Inside that safe edge, staff were free to get going themselves. Rather than watching a demo, they put their own work into it and saw what came out.
Introducing AI in a tightly regulated setting works, as long as you wall it off and start small. And adoption comes from people finding out in their own work what it does for them, far more than from any presentation about AI. Whoever has felt it needs no convincing.
Earlier at the same bank
Before AI was on the table, James worked at the same bank as a Lean Six Sigma Black Belt on Client Filtering: the process by which the bank runs clients past sanction and risk lists, part of know-your-customer (KYC). Every possible match is a hit that someone has to assess. That process was overflowing. James mapped it out and helped bring it under control.
What the diagnosis exposed
Figures from the process diagnosis itself. They describe the starting position rather than the end result: measure first, improve after.
The approach
The problem stated sharply on a single sheet: what prompted it, the problem statement, root causes, and actions with an owner and a date. Everyone sees the same picture.
The whole process drawn from start to finish, with a clear division of roles per step: who carries it out, who is accountable, who looks along. No work falling between two stools.
An MI dashboard that makes intake, backlog and quality visible. Steering on facts in place of gut feel.
The result: process mapped, causes named, steering in place.
From that point on the bank could improve deliberately, step by step, on the basis of figures. That is what in control means: you know what comes in, where it gets stuck and who owns it.
What this gives you
The temptation to throw software or AI at a problem straight away is strong. Automate a messy process, though, and you get mess at higher speed. That is why every Bravio engagement starts with the Blueprint: map out how you work and where it leaks away, and build after that.
Walled off, small, and inside their own work. That is how we set up your AI layer too: the AI handles low-risk, reversible work itself, and everything that goes out sits ready as a draft and leaves only with your approval.
Client acceptance and screening largely automated, with a 65% shorter lead time.
Read the case →What we build for agencies, firms and advisers: one workspace for clients, quotes and follow-up.
View the sector →What already works today, and what is hype. A level-headed overview without the sales talk.
Read the article →