Synthetic Data and Agentic AI

Banks are testing products on fake customers. It’s fast, cheap and ethically questionable.

Financial institutions are quietly replacing real customers with algorithmic clones to circumvent stringent data privacy laws and accelerate time-to-market.

Testing a new credit card or AI investing app traditionally takes months of testing. For bank product developers, the synthetic consumer, who never sleeps or complains to regulators, and costs a fraction of a penny to interview, represents a fast, highly attractive alternative, leading to its adoption across the industry.

The US bank deploys synthetic audiences to model consumer segments such as high-net-worth households, and tests messaging and refines campaigns before launch. Regulatory sandboxes encourage this practice to keep pace with AI-powered fintech. Barclays, Lloyds Banking Group and UBS UK are part of the FCA’s AI Live Testing initiative, which uses advanced AI systems to test products and simulate market stresses.

NatWest, Monzo and Symantec, meanwhile, explore synthetic data ecosystems to train AI models, while JPMorgan Chase generates synthetic financial data to simulate market behavior for risk management and product design.

Fast adoption, zero governance

Industry experts warn that the real challenge is balancing the speed of agentic AI with the need for strong governance.

“Most banking leaders believe agentic AI can advance rapidly if governance is not seen as a barrier. But in practice, governance is what makes these systems deployable at scale. A key part of this is robust testing against representative ground truth, and synthetic data provides a powerful proxy that enables banks to stress-test products against rare scenarios and marginal cases,” said EY Americas Financial Services Consulting AI. Practice leader Mudit Gupta said.

“The trade-off,” he adds, “is privacy: synthetic data is often considered inherently secure when it can still leak sensitive signals through inference and linkage risks. It can also replicate and scale historical biases, embedding them behind a layer of abstraction that makes them harder to detect, audit, and challenge – turning a governance shortcut into a long-term ethical risk.”

Ultimately, the rush to deploy synthetic consumers provides undeniable momentum, but the industry will quickly have to confront whether these powerful proxies – if not rigorously governed – will serve their purpose as a testing shortcut or simply institutionalize Wall Street’s next major ethics crisis.

This article is published in the June 2026 issue Global Finance Journal.