U.S. Bank: AI synthetic audiences for its Power of Us campaign
U.S. Bank and agency Supernatural AI used AI synthetic audiences to shape and test its Power of Us campaign, reaching market in under four months.
U.S. Bank worked with the agency Supernatural AI to build AI-generated “synthetic audiences” of its target customers, then used them to set the strategy and test the creative for its 2024 national brand campaign, “The Power of Us”. The project went from start to market in less than four months.
At a glance
| Item | Detail |
|---|---|
| Company | U.S. Bank (U.S. Bancorp) |
| Industry | Finance |
| Use cases | Market research, campaign strategy, creative testing |
| Functions involved | Marketing |
| Reported results | No return-on-investment figures disclosed; the agency says speed to market was under four months; the CMO says themes overlapped 90% to 95% with tests on real people |
| Moments rating | B (reference) |
| AI scenario score | 72 |
Case details
Background
Michael Lacorazza joined U.S. Bank as CMO in 2023 and, he told AdExchanger, wanted to rethink the bank’s brand storytelling and go-to-market approach. The usual route of market research and focus groups, followed by feeding the insights into creative production and media planning, could take six months or more. The bank wanted to move faster and bring in fresh thinking.
What they did
The bank partnered with Supernatural AI, a New York City startup. According to release details reported by Marketing Dive on 3 June 2024, the agency used AI to create avatars of six core audiences: young affluent, midlife affluent, high net worth, small business owners, corporate and commercial. It used them to develop the baseline campaign strategy, “powering potential”, and then to test that strategy and the first creative executions.
In a June 2025 interview, Lacorazza described five audience profiles, assembled from multiple third-party data sources and trained on different models. The team questions them through a large language model chat interface, much like a focus group. That includes topics too sensitive to put to people directly, such as how couples divide their finances.
Results and data
The agency says every AI audience model understood the double meaning of the line “The power of US”, and it used that finding to keep the campaign consistent across all elements. Speed to market was under four months. The campaign ran across connected TV, broadcast, out-of-home, digital, social and sponsorships. Lacorazza says tests with real people returned themes with a 90% to 95% overlap, and that humans still make the final decisions. No business results for the campaign have been published.
What to take from it
Synthetic audiences are useful for narrowing directions and checking how a message lands before formal research. Validate them against a real sample first, and re-check as the models change. Separately, American Banker reported in January 2026 that U.S. Bank had added a generative AI assistant to its developer portal to help business clients and partners find and implement its APIs.
Limits
The 90% to 95% figure comes from the CMO in an interview; the test method and sample size are not public. The under-four-months figure comes from the agency’s release. The two reports disagree on the number of audiences (six versus five). Synthetic audiences are only as good as the third-party data behind them and do not replace measurement of real customer behaviour.
Sources
Original source: Marketing Dive · U.S. Bank taps AI to create audience models for national campaign (2024-06-03) Additional sources: AdExchanger · US Bank Is Using Synthetic Audiences To Generate Real Customer Insights (2025-06-25); American Banker · U.S. Bank gives embedded banking clients a gen AI assistant (2026-01-15)
This is an industry research entry: a publicly documented third-party AI case, not a Moments client case. For projects Moments delivered itself, see client cases.