Date:
An introduction to a six-part series on bringing AI into regulated financial processes
Alyona Bulda and Marina Kudriavtceva, Exactpro
A magician who knows exactly what Aladdin's old lamp can do comes to the palace in disguise:
“He bought a dozen copper lamps, put them into a basket, and went to the palace, crying: ‘New lamps for old!’ followed by a jeering crowd.”
An old lamp sits on a cornice in the hall. No one in the palace remembers what it does:
“The princess, not knowing its value, laughingly bade the slave take it and make the exchange.”
“Aladdin and the Wonderful Lamp”, in The Arabian Nights' Entertainments, ed. Andrew Lang (1898)
The offer at the gate
Every financial institution is now hearing the magician's cry. New tools arrive weekly, each promising to do in seconds what a team does in a day. The offer is often a good one: ‘AI can take friction out of work that people find slow and repetitive, and firms that ignore it will fall behind’.
But the princess's mistake was not in taking a new lamp, it was in giving away the old one without knowing what it did. In financial infrastructure, the old lamp is everything an established process quietly provides: controls refined over years, checks that catch errors before they reach the market, the judgement of the people who run it, and behaviour that has been tested until it can be trusted. Swap this process for something new without understanding those aspects, and the loss only shows when it is too late to reverse.

Why this, and why now
The question facing firms is no longer whether to use AI. In the Bank of England and FCA's latest survey, 86% of responding firms reported using it (Bank of England, FCA, 2026). The harder questions are where to put it, how to get it out of the pilot stage, and how to prove that it is safe in systems where a single error can be felt across the whole market.
Those are testing questions as much as technology ones. At Exactpro, we spend our working lives testing the systems at the core of financial markets, and AI testing is now a growing part of that work: testing the AI systems being built into those markets, and training teams to do it themselves. This series is our attempt to share what that work teaches: it is practical, evidence-based and written for the people who have to make these decisions, not for the people selling the lamps.
What we will cover
- Careful What You Wish For: where AI fits in a financial firm's processes, and where it doesn't.
- Out of the Jar: why most AI pilots never reach production, and what the ones that do have in common.
- A Different Story Every Night: how to test a system that doesn't give the same answer twice.
- Open Sesame: data and governance: who holds the key, and what has to be in place before you start.
- The Porter and the Sailor: how teams, skills and roles change when AI enters a process.
- The Caliph's Night Walk: what regulators and supervisors expect, and how to be ready when they look.
Each article opens with a scene from the Thousand and One Nights, turns to one practical question, draws on published research and practitioners' views, and ends with a one-line takeaway for the board.
The tales are a reminder that people have been bargaining with powerful, unpredictable helpers for a very long time. The ones who did well were rarely the strongest or the luckiest. They were the ones who asked the right question before they made the exchange.