Testing a new credit card or banking product requires months of regulatory vetting and customer recruitment. Now banks are building the customer instead.
Financial institutions are replacing real customers with AI-generated surrogates – synthetic profiles that cost almost nothing and carry none of the compliance exposure associated with real customer data, Global Finance I mentioned. The artificial consumer doesn’t just compress timelines. It changes how banks bring products to market.
Adoption is spreading across major institutions on both sides of the Atlantic. Global Finance reported that the US bank is deploying synthetic audiences to model consumer segments such as high-net-worth households, test messaging and optimize campaigns before launch.
JPMorgan Chase creates synthetic financial data to simulate market behaviors to manage risk and design products. NatWest, Monzo and Santander use synthetic data ecosystems to train AI models.
The FCA is conducting live tests of AI but governance questions remain
In the United Kingdom, the Financial Conduct Authority (FCA) has moved to bring this practice into a regulatory framework. The FCA’s AI Live Testing Initiative launched its first batch in October, including NatWest, Monzo and Santander. The second group began in April, adding Barclays, Lloyds Banking Group, UBS and the Financial Conduct Authority (FCA). Announce. Use cases include agent payments, anti-money laundering detection, and know-your-customer checks. The testing will conclude by the end of 2026, with the evaluation report due in the first quarter of 2027.
The authority described the initiative as the first of its kind in the agency’s financial sector I mentioned. The companies cited direct testing of AI as a way to overcome what they called “proof of concept paralysis,” where AI initiatives stall due to regulatory uncertainty.
Advertisement: Scroll to continue
But the sandbox has limits. Most banking sector leaders believe that agentic AI could move faster if governance was not seen as a constraint. Mudit GuptaArtificial Intelligence practice leader at financial services consultancy EY in the Americas, told Global Finance. “But in practice, governance is what makes these systems deployable at scale,” Gupta said.
Gupta added that synthetic data is often treated as inherently secure. not so. It can leak sensitive signals through inference and correlation risks. It can also replicate and extend historical biases, embedding them behind a layer of abstraction, making them difficult to detect, review and challenge, he said.
Synthetic data extends to treasury operations and fraud detection
The scale of adoption makes the issue of management urgent. AS PYMNTS I mentionedHowever, the technology is already moving into treasury and finance operations, where forecasting models have historically relied on data that quickly becomes outdated.
Regulatory bodies are unlikely to treat AI outputs as abstract or low-risk. Fraud by an unauthorized party accounts for 71% of incidents and losses at financial institutions, due to credential theft and account takeovers, PYMNTS I mentioned. These are areas where AI is being deployed to make real-time judgments about identity, authorization and intent. The Financial Conduct Authority (FCA) said it will publish a report on good and bad practice around artificial intelligence in financial services later in 2026.
For all of our PYMNTS AI coverage, subscribe to our daily newsletter Amnesty International newsletter.
