Report

Speed Is a Revenue Strategy

Every large bank bought AI. Almost none of them can point at what changed. New research from Cornerstone Advisors maps where AI agents actually stand in bank marketing — and what separates the institutions getting a return from the ones getting output.

Research by

Cover of the Cornerstone Advisors report Speed Is a Revenue Strategy: How AI Agents Accelerate Bank Marketing

Get the report

Enter your email and we’ll send you the PDF.

The full 26-page report, straight to your inbox. We’ll send you the occasional thing we think is worth your time. Unsubscribe whenever.

001

Everyone bought the same AI.

Generative AI is now standard in bank marketing. Agents that can actually decide or act on something are not. The gap between those two numbers is wider than the maturity curve explains.

63%

use generative AI to create content

use AI agents that can execute an action

002

You already know where the value went.

Asked where agents would have the biggest business impact, marketing executives put orchestration and offer recommendations first and second. Content generation came third.

Both of the top answers are decisions, not drafts. The industry has worked out that the value moved to the choice. The spending hasn't followed it.

27%

campaign & journey orchestration

25%

offer & product recommendations

17%

content generation

003

Speed pays. Nobody counts it.

Speed to market came last among the performance factors bank and credit union marketing leaders say they measure. That is a gap in the reporting line, not a verdict on its value — the revenue is real whether or not anyone is counting it.

But what doesn't get counted doesn't get funded. That single fact explains why the tools that deliver speed sit unbought while opportunity windows close in days and review cycles run four to six weeks.

47%

call agentic AI experimental or emerging

20%

haven't discussed it at all

call it a top strategic priority

63%

expect to allow autonomy within two years

004

The bottleneck was never the writing.

Drafting is roughly a tenth of the elapsed time on a campaign. Review owns most of the rest. Every hour AI gave back came out of the small part — which is why faster writing has done so little for campaign velocity.

Institutions that read the queue as a risk-appetite problem add process. Ones that read it as a capacity problem add headcount. Neither changes the arithmetic. The ones making progress treated compliance as throughput and moved review into the work instead of after it.

10%

of campaign time is actually spent drafting

4–6

weeks of compliance review, on a good day

005

What to require before you buy.

The report's sharpest advice is to stop asking vendors what their system does and start asking how it decides. Three questions do most of that work — and most institutions can't answer the third about tools they already own.

What was the model that does the deciding trained on? Not the model that writes — the one that predicts. A general-purpose model has never seen your performance history, and that data was never in its training set.
Does a score come with a reason, and does it say when it isn't sure? A number on its own is an assertion. A model that flags its own low-confidence cases is more credible than one that always answers.
Does the loop ever close? Send what was ranked, then go back and check. Without that, a prediction is something you are asked to take on faith.

About the research

Who was surveyed

Cornerstone Advisors surveyed marketing executives at banks and credit unions on what they have deployed, what they expect to be comfortable with in two years, and where they believe agents will matter most. Cornerstone has spent more than 20 years advising financial institutions on strategy, technology and performance.

The research and its conclusions are Cornerstone's. Persado commissioned the study.

AuthorsRon Shevlin, Abbie Jones, Elizabeth Gujral

RespondentsBank & credit union marketing executives

PublishedSeptember 2026

Commissioned byPersado

From Persado

Everyone can make the content now. Almost nobody can tell you which one to send.

The cost of producing a marketing message has fallen to roughly zero. The value moved to selection — predicting how a message will perform, ranking the options, and improving the ones that fall short, before anything reaches a customer.

That's what Persado does. We don't care what AI you use. Every model, agency and in-house team becomes a source of candidates. We score each one for predicted performance and for compliance in the same pass, and hand back your book ranked by what to fix first.

8 of 10

largest U.S. banks work with Persado

2B+

messages analysed

120K+

performance-labelled campaigns

20+

regulatory frameworks built into scoring