FinTechTalk: Rethinking financial crime decision-making in the age of scalable AI workforces

Views on news 

Financial fraud has expanded into a high-volume global activity. According to Vyntra’s 2026 report, nearly two-thirds of scams succeed within a day of first contact, leaving little opportunity for intervention once engagement begins. Generative AI appears central to this acceleration, reducing the time required to assemble convincing phishing campaigns from more than 16 hours to under 5 minutes.

 

The report outlines a wide mix of fraud types, including executive impersonation, phishing-led account takeovers, and recruitment scams, all increasingly supported by AI-generated content. Only two years ago, you had to go to a consultancy and pay millions of dollars to get the software that gen AI can create you in an hour.  

 

How can experts keep up with technological development? 

Financial institutions must play catch-up with criminals while being compliant and having robust governance frameworks in place. Therefore, they must get more agile in a controlled way. Agentic AI can offer the scale that enables them to take on criminals relying on gen AI and other advanced technologies – they can manipulate bigger data sets, take a much broader set of signals and mesh them together.

 

AI is a great tool for identifying fraudulent transactions, but false positives have been a pain point for a long time. Advanced AI can now be leveraged for reducing the number of false positives significantly but, at the current stage of AI development, it’s key that a human stays in the loop to get viable, explainable and compliant outcomes.   

 

However, there are now more than 20 large deployments with global tier-one institutions executed by Silent Eight, a start-up, where decisions are made autonomously in relation to sanctions, transaction monitoring and customer due diligence. Gen AI is completely reshaping the landscape, including the job of the financial anti-crime professional – they need to know now how to develop AI agents and must keep up with technological advancement acquiring new capabilities. And as the number of deployments increases within a jurisdiction, the argument that regulation stands in the way of implementing these technologies will no longer hold.

 

AI is not a silver bullet, though. You must have a clear idea about what you want to achieve with it. You must also consider the maintenance and updating of the model when choosing one, as well as whether these activities will be the vendor’s responsibility and what price tag this ongoing service comes with. Continuous self-education is key for experts dealing with AI.  

 

The panel’s advice 

• The AI-driven data science part of the financial job will most probably disappear in 12-15 months.  
• Adoption of agentic AI is gathering pace with businesses actively seeking deployment opportunities.  
• If you don’t refine an AI model every 5-6 months, its efficacy will fall.  
• To stay up to date, get an understanding of what gen AI and agentic systems are – use Coursera, Data Camp. 
• Experiment outside work to learn what the boundaries of safe and secure AI use are.  
• Get a Mac Mini and deploy an OpenClaw bot to learn how you can work with it, as well as its limitations. Experiment with the full gamut of other AI tools’ functionalities too.