Implementing AI Transformation within a Central Bank – Aizhan Arinova [Season 3 | Episode #15]

Tom Mitchell 00:00:33 Welcome to the P.I.T. Exchange. I’m Tom Mitchell from Currency Research, and today I’m joined by Aizhan Arinova, who is Director of Digital Transformation at the National Bank of Kazakhstan. Aizhan, thank you very much for joining me.

Aizhan Arinova 00:00:47 Thank you, Tom. Thank you for your time.

Tom Mitchell 00:00:50 We’re recording this today against the backdrop both of the Silk Road Cash and Payments Conference that’s taking place in Almaty, Kazakhstan in a couple of weeks’ time, but also against the backdrop of Kazakhstan’s 2026 Year of Digitization and AI. I want to use our time today to get into what that actually means in practice at the National Bank — things like strategy, process, platforms, data, and the results you’re seeing.

Let’s get right into it. Could you introduce your role and how the Digital Transformation Department fits within the National Bank of Kazakhstan?

Aizhan Arinova 00:01:29 Thank you, Tom, for the question. I’m the Head of the Digital Transformation Department, and I’m responsible for the whole digital agenda within the National Bank of Kazakhstan. In December 2024, we launched our digital strategy — a five-year strategy with more than 100 projects. We’re now in the second year of implementation. As a result of changing our operational model, we launched the new Digital Transformation Department last year, in June 2025, and I joined the team as part of that.

Our strategy has five pillars, and our goal is to build a modern central bank — not only in our country, but across the Central Asia region. I think we have all the resources and the timing is right to reach this goal.

Aizhan Arinova 00:02:51 The first pillar is building national financial digital infrastructure. The second is building new, modern IT infrastructure. The other pillars are building data governance and building digital platforms. The fifth pillar is cultural transformation — because without people who are ready to work in a new digital era, we can’t build a digital central bank.

This year, as part of the broader government agenda on artificial intelligence, we’re more focused on implementing AI-related projects — I’ll go deeper into that later. We have 38 projects to deliver this year.

Tom Mitchell 00:04:22 A fantastically busy year. Thirty-eight projects for delivery this year, out of the roughly 100 projects across the full five years. That’s a lot going on — and a very ambitious strategy, which is great to see.

Looking at the digital transformation strategy overall — what are the current core priorities? What are you working on right now, out of those 38 projects? And on governance: who sets the mandate for this strategy, and how is delivery rolled out across the bank’s different departments?

Aizhan Arinova 00:05:18 Let me start with the second part. As part of our new operational model, there’s a new role — Deputy Chairman, which we call the Chief Digital Transformation Officer. That’s a new line, and he’s primarily responsible for the whole digital transformation agenda.

We also have a clear mandate from our Governor, and his political backing is helping us move quickly. We’ve built two operational vehicles for the digital transformation agenda. The first is what we call the Value Delivery Office — a project management office. Every week we discuss different projects and situations with the Deputy Governor. The second vehicle is a technical consultation office — an architectural body.

Those two vehicles help us move quickly. Another idea we’ve been developing since the start of the strategy: different departments within the bank each own different business processes and systems, and we’re building partnership relationships with them.

Aizhan Arinova 00:07:32 We’re developing technical solutions together with those departments — it’s a genuinely warm partnership relationship with our internal business units. Another thing that helps us move quickly: every quarter we hold a Quarterly Business Review, a big event where we show results for that quarter — what’s happening across the digital agenda, the status of strategy implementation, blockers, workarounds, and so on. We gather top management and key stakeholders and work through issues with them openly.

Tom Mitchell 00:08:39 Fantastic, thank you for that overview — a lot to take in there, and I appreciate you outlining how the bank is approaching this. Let’s look a bit closer at how core internal processes and workflows are being digitized, and the change management involved. You’ve said you’re working closely with other departments. Can you give an example of a process at the National Bank that’s been substantially digitized recently — what was involved, and what were the challenges in rolling it out?

Aizhan Arinova 00:09:27 There are more than 30 departments within NBK and more than 25 IT systems. Every system and department has its own mandate and function. Last year we focused on foundations — solving quick problems and working on infrastructure, some fairly simple solutions. This year we’re already seeing results from a bigger project we call the NBK Data Factory. We’ve now built the capability and the momentum to work across a wider range of agendas with different departments.

Aizhan Arinova 00:10:44 Our approach to change management is built around quick wins alongside systematic solutions. Last year we invested heavily in the NBK Data Factory project — let me go deeper on that. It’s a technological platform that helps our analytical departments use and manage data. We reduced time-to-data and time-to-analytics by four to five times. That was a big win, and it showed our partner departments that we could take on other projects too, so we started working on other processes and systems as well. From my perspective, Data Factory was the first big win.

It’s not finished — it’s a three-to-five-year project — but we already have concrete results, solving technical and analytical problems for the business. As a result, we’ve earned support and trust from those departments. Right now we’re modernizing our IT landscape, and this year we started using “NBK AI-assisted coding” as an approach — trying to build many systems from scratch using AI-assisted coding. It’s reduced development time drastically: what used to take six months now takes a maximum of one month — six to ten times faster. As a result of this AI shift, I think we’ll finish our strategy faster than planned.

Tom Mitchell 00:13:57 The Data Factory work is presumably part of what you’ll be presenting at the Silk Road Cash and Payments Conference — I think that’ll be an eagerly anticipated session, where we can really dig into how AI is being implemented across the bank’s systems.

The bank has publicly said that within two to three years, it expects nearly all software at the bank to be built with AI assistance by default — largely for the speed reasons you’ve described. It’s also been reported that in 2026 the bank is anticipating bringing in more than 20 practical AI-based solutions, including AI agents for anti-fraud investigations, financial supervision, currency monitoring, knowledge management, and software development. Looking at the digital platforms AI is helping to build — which are the most transformative ones you expect to roll out over the next couple of years? And are these being developed primarily in-house, or are you also working with technology partners?

Aizhan Arinova 00:15:27 Let me start with the NBK AI platform. This year we launched our AI implementation strategy — a two-year strategy, though things are moving so fast we’ll probably revisit it next year. As part of that strategy we built the NBK platform, launched in January this year — a fully independent, entirely in-house platform, built on our own local GPU cluster. Right now we have more than ten AI agents and services running on it.

We’re working on both horizontal and vertical adoption of AI. Horizontally, we’re building everyday services for the whole of NBK.

Aizhan Arinova 00:17:06 By “whole of NBK” I mean our central body in Astana, plus 20 regional branches — more than 2,000 people in total — and more than ten subsidiaries that we share services with, because we want to build an AI ecosystem where people can share AI services across companies and teams. Today we have operational services like a translator and a tool for transcribing audio and video, our own RAG system built on internal documentation and knowledge bases, and various tasks and skills people can use within the MBK platform. We track metrics like daily and monthly active users — right now around 30–35% of people use the platform daily.

Aizhan Arinova 00:19:01 By the end of this year our goal is 80% adoption across NBK. We want the platform to become a kind of digital workplace — a starting page everyone in the bank can use.

On the vertical side, we have standalone projects and services built for specific needs or departments — for example, anti-fraud solutions, transaction analytics, or work within specific data domains. The platform integrates with Data Factory, and we have a text-to-SQL solution so people without SQL knowledge can work with data. We have 19 data domains, and we’re building AI agents specific to each one, since each has its own requirements — that’s our vertical adoption of AI.

We’re also already using the AI-assisted coding approach I mentioned — our developers and data engineers use our internal GPU cluster and internal models to improve speed and quality of development; that’s the professional use of AI. And there’s a third layer — self-service AI — where anyone without development or IT knowledge can build their own service or app using a sandbox on the MBK platform. We’re working on scaling that up. As part of this, we recently ran the “MBK Challenge” over two months: people built their own apps or solutions using the sandbox, with support from our team. We started with 103 ideas; within eight weeks, 27 solutions had been built; in the end, eight teams pitched their solutions to NBK top management. It was a great boost to AI adoption — right now it feels like everyone at NBK is interested in AI and these tools. Going forward we want regular hackathons and “datathons” as part of change management — experimenting with different ideas through events. It’s been a real success story.

Tom Mitchell 00:24:33 It’s fantastic to see so many colleagues engaged with AI. My own experience was being told for a long time that it would be transformative, and once I actually started using it, I couldn’t imagine going back — there’s a real tipping point once people start using it a bit. As all these systems are developed with different teams contributing, there’s a big question of interoperability. How do you make sure these systems work together, both within the National Bank and within the wider financial system?

Aizhan Arinova 00:25:22 We’ve built a systemic pipeline where every layer and level of development has checkpoints and guardrails, so I’m confident everything works properly. There’s a cybersecurity team and a technical team involved throughout the pipeline — everyone is responsible for their part of the code or the application. As a result, we end up with solid solutions.

Tom Mitchell 00:26:15 Perfect, thank you. Moving on to data — obviously a critical part of any financial transaction, and data management especially. Looking at the infrastructure and governance underpinning digitalization at the National Bank — how has the bank’s approach to collecting, managing, and using data evolved as part of this transformation?

Aizhan Arinova 00:26:49 Let me start with data governance, because solid data governance is essential — true adoption of AI isn’t possible without a correct “golden” semantic layer describing the different data domains; without that, you can’t move quickly on the AI agenda.

Second, as part of our data governance operational model, we’ve built a Data Governance Office, headed by another Deputy Governor — Magzhan Berdaliyev — who oversees the analytical department and is the main sponsor ensuring Data Factory and data governance work well. We divided our data landscape into 19 domains and are working through them step by step, starting with the domains that have the most users and the greatest need.

Aizhan Arinova 00:28:55 On the data management side, in 2025 we started building a modern data warehouse, which we’re still working on — but we’ve built the foundation over these two years. Right now we’re more focused on master data management and data governance — describing data objects within the warehouse: data marts, data models, dashboards, and so on. As part of master data management, we’re aiming to finish this project this year, working toward “golden truth” and golden records across different data sources. As a result, we’re already seeing reduced time-to-data and time-to-analytics for our analytical department. We’re also developing our data governance policy, because without a clear methodology this won’t work in the long run.

One more point: alongside self-service AI, we have self-service data analytics — our analytical department building their own dashboards and data products on our technology stack. Part of our approach is building expertise within business users in the analytical departments, so they can work with these tools without needing the technical team involved every time.

Tom Mitchell 00:32:08 Looking at team composition and data management more broadly — do you think there are data challenges that are unique to central banks that other financial institutions don’t face?

Aizhan Arinova 00:32:30 There’s only one central bank per country, so you don’t always have a direct reference point within your own country — sometimes there’s no one to compare notes with. That said, we do have horizontal relationships with data teams from Uzbekistan, Kyrgyzstan, Russia, and elsewhere in Central Asia, though the legal and regulatory context differs. That’s one obstacle for central banks, but I don’t see any that are truly crucial — it always comes down to resources, and thinking through the project like any other.

Sometimes, as a fairly conservative government body, especially in the first year of transformation, there were some emotional obstacles from different departments. But now we’ve built genuine partnership relationships with them, and they’re working with us. That wasn’t specific to Data Factory or AI projects — it applied across all projects, simply because we’re a conservative government body with a certain level of bureaucracy and specific business processes. I wouldn’t call them obstacles exactly — just the specifics of how we work.

Tom Mitchell 00:35:22 Since we’re partway into the five-year plan — you mentioned we’re in year two — what efficiency gains have you seen so far? What have been the quickest, easiest wins from the work you’ve done?

Aizhan Arinova 00:35:57 That’s a hard question, because we’re not a commercial bank — you can’t measure efficiency gains purely in terms of money or revenue. Sometimes the gains are more qualitative. We use an NPS-style approach, asking departments how their processes and systems are improving. But it’s difficult, because we didn’t have time to establish baselines for every process and system before we started.

Aizhan Arinova 00:37:14 We didn’t have existing documented business processes, methodology, or policy for many areas, so measuring specific metrics across every project and process is hard — it’s practically impossible to measure everything, and we don’t have the time for that. As I mentioned, we did measure specific things, like reduced time-to-data for certain domains, but not across the whole system.

When departments come to us with requests, the Value Delivery Office’s main question is: what value will this deliver? Are we reducing hours people spend working in the system, reducing steps in a process, and so on. It varies case by case — I don’t have one overall number or percentage for the whole system, but we are trying to measure it. It’s difficult for a government organization.

Tom Mitchell 00:39:13 That’s understandable — it can be hard to put a single number on projects like this. But you’ve pointed to real early wins, like reducing system rollout time from six months to one month using AI — a great example of what’s working. Looking ahead, since you and your colleagues will be discussing this in much more depth with attendees in Almaty in a couple of weeks — for people coming from outside Kazakhstan or the region, what’s the one thing you’d hope they’d understand before they arrive?

Aizhan Arinova 00:40:34 That’s a difficult question. First, thank you for inviting me to be part of the Silk Road Cash and Payments Conference — I’ll be there, and I’ll go deeper into our projects related to currency and our various initiatives, and into our AI agenda. Hopefully we’ll be able to show some new applications and services we’re currently working on.

For people coming from other countries — it’s an honor to host our guests in Almaty and in Kazakhstan. We’re very open to discussion, and during or after the conference we’re inviting everyone to visit our new building in Astana — there may be opportunities for collaboration with people attending. We hope the agenda is interesting and the policy discussions are deep and productive for everyone. Please come and join us.

Tom Mitchell 00:42:20 Thank you — and we’re very grateful to be working with the National Bank of Kazakhstan on this event again. Thank you for joining us today to give listeners a preview of what attendees will hear.

At the event in September, the National Bank of Kazakhstan team will present on topics including AI-driven data transformation as part of modernizing and integrating cash and payment systems, and separately, a spotlight on safeguarding digital payments through their anti-fraud center. Attendees in Almaty will hear much more depth than we’ve been able to cover today.

Of course, the Silk Road event isn’t only about payments and digital topics — cash remains central to the agenda too, and we’re looking forward to working closely with the National Bank of Kazakhstan’s cash department. Across three days, the event has nine main-stage topics, five additional spotlight and technical innovation sessions, plus keynotes and programming covering everything from cash’s strategic role in modern economies, through design and security, to infrastructure resilience and crisis preparedness. There’s also a dedicated spotlight on Kazakhstan’s own currency production, with both the Banknote Factory of Kazakhstan and the Mint of Kazakhstan taking delegates through how the country manufactures its banknotes and coins — running alongside the closed-doors Central Bank Summit this year, focused on cash and payment system resilience across Central Asia.

We have a strong lineup of presenting organizations sharing the stage with the National Bank of Kazakhstan this year, including the Reserve Bank of India, the Central Bank of Mongolia, Banque de France, the Royal Monetary Authority of Bhutan, the Central Bank of Egypt, the Swiss National Bank, and the Central Bank of Azerbaijan, alongside industry names such as CPS, ECM, Crane, GMD, Travelex, and Sargent & Greenleaf. Anyone wanting the full agenda can visit the Silk Road Conference website.

What’s especially exciting this year is seeing how many central banks from across Central Asia and the wider region are coming together in one room, alongside much of Kazakhstan’s own banking sector and colleagues from neighboring countries. Beyond the sessions themselves, the National Bank of Kazakhstan is kindly hosting a cultural tour of Almaty, a delegate dinner, and an exclusive tour of production facilities for central bank guests.

To everyone listening — the National Bank of Kazakhstan will be at the Silk Road Cash and Payments Conference in Almaty from the 15th to the 17th of September, 2026, and we’d love for you to join us. Thank you for joining us today, and thank you to all our listeners for tuning in to the Exchange. We look forward to seeing you next time.

Aizhan Arinova 00:45:40 Thank you, Tom.