Only 9% of bank and credit union marketing teams use AI agents that can act

Among marketing teams at banks and credit unions, 63% use generative AI to produce content. The share using AI agents that can carry out an action is 9%.

Both numbers appear in Speed Is a Revenue Strategy: How AI Agents Accelerate Bank Marketing, a new report from Cornerstone Advisors that Persado commissioned. The report draws on a recent study of marketing leaders at banks and credit unions. A third number falls between the other two: 32% of teams apply AI to analytics and decision support. That puts adoption at its highest in the step that generates words, with use dropping off in the steps that decide and act.

Cornerstone's argument is that institutions judge marketing by campaign results, engagement and personalization, and rarely by how long it takes to get from a customer signal to a compliant message in market.

The four-to-six-week approval queue

Compliance review ranks among the biggest obstacles to faster execution, the study found, with cycles at some institutions running four to six weeks. Customer needs and competitive openings, by the report's account, can come and go within days or hours.

At institutions where compliance review runs four to six weeks, a campaign drafted by generative AI waits in the same queue as one drafted by hand. That's the practical reading of the 63% figure. Content tools speed up the step before the bottleneck, and the bottleneck still sets the pace.

Persado paid for the research. The company sells a Marketing Compliance AI platform for marketing and legal teams in financial services, which combines regulatory compliance analysis, performance prediction scoring and brand-fit insights. Emmanuel Richard, Persado's chief revenue officer, said the institutions best placed to capture deposits, loans, loyalty and wallet share will be the ones that go from insight to outreach in hours rather than weeks, while rivals are still waiting on approvals.

Comfort is running ahead of priority

Marketing leaders say they're ready to let agents act, but they haven't put agents at the top of the plan. The survey figures sit side by side like this:

There's a wide distance between 5% and nearly two-thirds. Leaders expect to hand agents real work within two years, yet almost none have made that work this year's top line item. Offer and product recommendations came a close second to orchestration at 25%, which puts the expected payoff in deciding what to send and when, not in writing it.

The report names four practical applications:

• Integrated compliance review, to validate content before it ships
• Intelligent customer journey orchestration, to trigger timely outreach
• Generative content optimization, to speed up testing
• Real-time segmentation, to keep customer segments current

For the compliance officer who would have to sign off on the "approved boundaries" in that two-thirds figure, defining them is most of the job.

Proving speed pays needs better data plumbing

Elizabeth Gujral, a Cornerstone director and co-author, argues that institutions should treat speed as a growth measure, not only as a matter of efficiency. The strongest objection to that view comes from Cornerstone's own earlier research. Its report The Marketing ROI Gap in Banking was released April 28 and commissioned by Fintel Connect. It draws on responses from 126 senior executives at banks and credit unions in the U.S., and nearly 6 in 10 of them say the core or CRM platform they rely on restricts how well they can measure marketing ROI.

According to CU Today, none of the respondents said their institution could reliably attribute every one of the six outcome measures in Cornerstone's survey, which covered account openings, funded accounts, balances and profitability. To show that a campaign launched in hours earned more than one launched in six weeks, an institution needs exactly the kind of attribution that the finding says is missing.

For institutions whose core can't serve data in real time, the agent uses that depend on live signals, like real-time segmentation and journey triggers, would stall at the data layer before a compliance check ever runs. Gujral's point in the DCI report fits here: a core can hold years of transaction history, but if the bank can't reach it at the moment a customer is ready to buy or about to leave, it can't act on it.

The 5% figure to watch

The comfort expectation has a two-year window, and the 5% priority figure is what will show whether institutions are moving on it. For governed autonomy to arrive on the schedule respondents predict, agentic AI would have to climb well past 5% on the priority list, and the first place that shows up will be how bank marketing and compliance teams split their budgets.