Tieto Banktech, a banking software developer, has outlined how a combination of AI-driven detection, real-time risk assessment and human oversight is being deployed to intercept fraud before losses occur.
The company’s Financial Crime Prevention (FCP) division uses AI models trained on large and complex datasets in real time, enabling them to identify patterns, assess risk and flag suspicious activity at an early stage. FCP gathers and analyses fraud patterns across the majority of Norwegian banks and a significant portion of the Swedish market, with insights shared between institutions to strengthen collective defences.
Central to FCP’s latest developments is Fraud Explore, a solution built on the company’s own Large Financial AI Model (LFM). Designed to bolster real-time fraud detection across the Nordics, with international expansion in scope. It
Where traditional rules provide a stable foundation, the AI layer identifies subtle behavioural signals that rules alone would miss.
In card fraud specifically, FCP operates a layered model. The first layer, Token Enrolment Monitoring (TEM), intercepts attempts to add payment cards to digital wallets, stopping between 70 and 80 per cent of token fraud attempts before the cards are ever used, it said. The second layer, Card Transaction Monitoring (CTM), analyses card transactions in real time, building a picture of normal spending behaviour for each individual customer and flagging deviations to reduce both fraud and false positives.
Sitting above the core financial model is a set of LLM-based AI agents that form FCP’s next generation of case handling. These agents are capable of structuring complex information, proposing assessments, automating repetitive tasks and delivering real-time decision support to case handlers, allowing investigators to direct their attention to situations requiring genuine human judgement.
Tieto Banktech’s head of Data & AI, Financial Crime Prevention, Nikolay Martyushenko said, “In fraud prevention, we use AI strategically – where it has the greatest effect. At the same time, people are always in control.”
Martyushenko added, “This investigative work takes time, but it is absolutely essential. If cases are not classified correctly, the models will learn the wrong things.”
He continued, “For us, fraud prevention is not just about algorithms, but about building, understanding and maintaining the right knowledge at all times.”
For more insights, read the full story here.
Copyright © 2026 FinTech Global
