Hi friends,

As you know, I'm currently in Kazakhstan, and I've been making a point of getting out to see what's actually happening in the local market from a financial-services perspective.

A few weeks ago, I went to the Central Asia Fintech Summit.

The first thing that struck me wasn't AI. It was how fintech itself is developing here. The banks are holding their ground. Venture investment into standalone fintechs is still relatively thin. And yet regulators and government are clearly trying to create room for new products, digital assets included, rather than waiting for the market to develop entirely on its own.

That tension kept coming up all day: move first, learn from what happens, then adjust.

One of the panels, The Real Economy of AI: From Pilot to P&L, polled the room live. Where does AI create the most real value in your organisation today?

Nobody said new revenue.

30% said cost reduction and productivity. 40% said better customer or public service. 7% said risk reduction. 23%, nearly a quarter of the room, said they're still mostly at the pilot stage.

I found that more interesting than another list of AI use cases. Productivity gains are visible. Service improvements are visible. The harder question is still how to connect both to a P&L line everyone agrees on.

But this isn't unique to Kazakhstan.

The clearest signal from the summit wasn't a panel. It was the posture underneath several of them: Kazakh regulators seem comfortable with an AI and tech-optimist approach, openly encouraging financial organisations to experiment, see what works, and develop the regulatory response alongside the market, or even after innovative solutions come to market, rather than waiting for regulatory certainty first.

That's quite different from the UK. But I don't think the comparison with the UK is as simple as Kazakhstan moves fast, the UK moves slowly. The FCA has its own AI Lab, including live testing and a Supercharged Sandbox, and its Mills Review, published in July and covered in Issue 11, explicitly looks at how AI could reshape retail financial services through 2030. It's also said plainly it doesn't plan AI-specific regulation, preferring existing, principles-based frameworks.

The difference I noticed was how comfortable the regulatory conversation seemed to be with learning through live market development in Kazakhstan, rather than trying to anticipate every outcome in advance.

One of the most revealing exchanges came from a panel on government data. The government-side argument was that public data should be put to work, not just stored, used as an economic asset.

The private-sector response brought it back down to earth. If connections to government data are already unreliable enough that customers are frustrated, monetising that data isn't the first problem to solve. Make it available. Make it reliable. Make it work in real time. Then build the business model on top.

Two very different vantage points, on the same stage. Government moving fast on ambition. The private sector still waiting for the basics to work reliably.

What execution looks like

Freedom Bank gave the clearest evidence of institutions already past the pilot stage: continuous-monitoring compliance agents flagging anomalies and routing them to a second layer of review, and a back-office and legal operation running 134 agents that took one process from two weeks to eleven minutes.

Unglamorous use cases: compliance, legal review, forms, fraud detection. Exactly where you'd expect AI to create measurable value first. Freedom isn't a fintech in the startup sense anymore, it's a bank, but it's building these capabilities in-house rather than buying a tool and calling it transformation.

If you do one thing before the next issue

Ask your own team the same question: where does AI create the most real value in your organisation today? New revenue, cost, productivity, risk reduction, customer service, or still mostly pilots.

Then compare your answers. You might be surprised by how close they are or how far apart.

Until next one,

- Gaziza