VIENNA, Va. — Is there a shortcut to getting member feedback? Financial institutions–including Navy Federal Credit Union–are increasingly turning to artificial intelligence-generated synthetic consumers and synthetic data to accelerate product development, improve marketing and conduct consumer research, with early testing suggesting the technology can complement—but not replace—traditional research using real people.
The growing interest comes as banks and credit unions seek ways to reduce development costs, protect consumer privacy and reach customer segments that are difficult to survey, according to two separate reports.
Among the institutions exploring the technology is Navy Federal Credit Union, which recently partnered with Qualtricson a pilot project comparing AI-generated synthetic survey responses with responses from actual consumers, Banking Dive reported.
Kathleen Myers, assistant vice president of research strategy at Navy Federal, said the technology could be particularly valuable for gathering feedback from active-duty military personnel, who often have limited availability to participate in traditional surveys.
‘Really Hard to Find’
“Active duty people are really, really hard to find,” Myers told Banking Dive. “The thought of being able — once you’re certain that the models behind it are kind of tuned appropriately — to use that for quick feedback or feedback at greater scale, I think is a huge opportunity for us.”
The pilot reflects a broader trend across the financial services industry.
According to research from Qualtrics cited by Banking Dive, 41% of market researchers already use synthetic data to supplement or replace human respondents, while 62% say they expect to use it in the future.
Meanwhile, Global Finance Magazine reported that banks are increasingly replacing portions of traditional consumer testing with algorithmically generated “synthetic consumers” designed to mimic real-world behaviors. The models allow institutions to test new products, marketing campaigns and digital services without relying on actual customer information, helping address privacy concerns while significantly shortening development timelines.
Testing a new credit card, mobile banking feature or AI-powered investment service has traditionally required months of customer interviews, research and compliance reviews, the publication noted. Synthetic consumers can be created and analyzed almost instantly and at a fraction of the cost.
How Pilot Test Worked
For its pilot, Navy Federal and Qualtrics surveyed 501 human respondents about trust and financial services related to a potential credit card offering, then generated 498 synthetic responses based on the results.
Researchers found the synthetic and human responses were remarkably similar overall, with average scores differing by less than 0.25% across survey questions.
Among the findings:
- Synthetic and human respondents generally reached similar conclusions.
- Synthetic respondents placed greater emphasis on efficiency and financial benefits.
- Synthetic responses were less likely to prioritize empathy and emotional considerations.
- Synthetic data appeared less susceptible to common survey biases, including acquiescence bias and social desirability bias.
Notable Differences
The study also identified notable differences.
When asked whether they were reluctant to apply for another credit card because they feared being rejected, 41% of human respondents agreed, compared with just 21% of synthetic respondents.
Myers suggested synthetic consumers may provide more candid responses because they are not influenced by embarrassment, shame or social pressures.
At the same time, synthetic respondents were significantly more trusting of various industries than their human counterparts and were much more likely to identify cybercrime as a top consumer concern.
According to the pilot, 24% of synthetic respondents cited cybercrime as a major concern, compared with just 6% of human respondents.
Difficult to Capture Emotions
Researchers concluded that while synthetic data often produces highly rational answers, it does not always capture the emotional and sometimes irrational factors that influence real consumer decisions.
“We’ve observed that synthetic data will offer more rational responses to questions we pose, navigating functional trade-offs logically and effectively,” Myers told Banking Dive.
Andy Pierce, global lead for value proposition innovation and design at Bain & Company, told Banking Dive that large language models frequently evaluate decisions through a highly rational lens, creating linear relationships between variables such as price and quality that do not always mirror actual consumer behavior.
However, Pierce said improvements in prompt engineering and AI training are helping systems better account for emotional factors such as brand loyalty and consumer preferences.
Looking Forward
Myers said Navy Federal envisions using synthetic data to test marketing messages, prioritize product concepts and improve survey design before conducting research with human participants.
“Breadth from the synthetics, but depth from the human,” she told Banking Dive, adding that traditional research remains essential when studies focus on emotions, identity, loyalty or other deeply human experiences.
What Banks are Doing
Banks are also expanding their use of synthetic consumers beyond survey research.
According to Global Finance Magazine, U.S. Bank uses synthetic audiences to model consumer groups, including affluent households, enabling the institution to test marketing campaigns before launch.
In the United Kingdom, the Financial Conduct Authority’s AI Live Testing initiative includes participation from Barclays, Lloyds Banking Group and UBS, allowing firms to evaluate AI-powered products and simulate market conditions before wider deployment.
Global Finance Magazine also reported that NatWest, Monzo and Santander are exploring synthetic data environments to train AI models, while JPMorgan Chase is using synthetic financial data to model market behavior for risk management and product development.
Not Keeping Pace
Despite the growing adoption, experts said governance and oversight have not kept pace with the technology.
“Most banking leaders believe agentic AI can move faster if governance weren’t perceived as a constraint,” Mudit Guptatold Global Finance Magazine. “But in practice, governance is what makes these systems deployable at scale.”
Gupta said synthetic data enables financial institutions to test products against unusual scenarios and edge cases that conventional testing may overlook. However, he cautioned that synthetic data is sometimes incorrectly viewed as inherently safe, even though it can still expose sensitive information through inference and linkage risks.
He also warned that synthetic datasets can replicate historical biases, potentially embedding discriminatory outcomes into future AI systems while making those biases more difficult to detect and audit, according to Global Finance Magazine.
