Uzbek Industrial and Construction Bank says it moved a multimodal AI platform from development into production in six months. Called SQB.NOVA, the in-house system lets departments reuse AI infrastructure and includes Uzbek speech recognition, a practical feature in a multilingual banking market.
Watch Desk analysis
What happened
The Times of Central Asia reports that Sergey Kozlov, the bank’s representative at the FINNEXT Asia forum in Tashkent, described the platform as an alternative to teams building separate systems for each task. The bank is aiming to make useful components, such as document processing and corporate-knowledge tools, available for reuse across departments.
Kozlov said SQB is still assessing the platform’s financial impact and prefers to discuss verified results rather than projected savings. He also identified credit-risk analysis and fraud detection as promising areas for AI in banking, not as proven results from this platform.
Why it matters
The interesting detail is not simply that a bank has put AI into production. It is the attempt to make deployments reusable, while building in a language capability that international products may not offer to the same degree. If the model works, each team may spend less time rebuilding the same foundations. The bank has not yet shown that this has reduced costs or improved business results, so the financial case remains a question rather than a victory lap.
The discussion also reaches beyond efficiency. Kozlov raised concerns about data exposure when staff use public AI services with banking documents, and about fairness and accountability in automated decisions. His position was that automation does not transfer responsibility away from the bank.
Our read
SQB.NOVA is a concrete example of a bank trying to move from isolated AI experiments to shared infrastructure, with Uzbek speech recognition as a notable local feature. That is more useful than another round of polished pilot demonstrations. But production is a milestone, not proof of impact: the bank says it is still measuring the financial results.
What to watch
- Whether SQB publishes measured results on costs, performance or business outcomes.
- Which reusable components other departments adopt, and how quickly.
- How the bank handles data security, explainability and human accountability in consequential decisions.
Discussion spark: Should banks prioritise shared in-house AI platforms that fit local languages, even before they can show clear financial returns, or should they wait for measurable results before scaling them?
Sources and evidence
- Source update (20 November 2026, 12:42 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.