On this page
A community bank or credit union with a few hundred employees rents its core system, the main ledger, from the vendor that runs it, and surrounds it with vendor products: online banking, card processing, loan origination (the software that takes a loan application through approval), document imaging. The work between those systems is done by people: a loan file checked against the application, an exceptions report worked line by line, a dispute assembled from a call note and three emails.
Those vendors are now adding AI to the products you rent, and the industry's surveys each report a share of institutions using it.
Neither tells you whether the hours your back office spends between systems went down.
Set each survey's figure next to the question it asked, and the answer turns on how narrow the question is.
About half of banks and credit unions have deployed generative AI, AI that writes text and answers, such as a chat assistant,2 and fewer than one community banker in ten says the bank is adopting AI across the whole institution.1
The number none of these surveys counts is the narrow one: how many pieces of your own work run on AI from start to finish.
About Half of Banks and Credit Unions Have Deployed Generative AI
Cornerstone Advisors, a banking consultancy, surveys senior executives at banks and credit unions of $250 million to $50 billion in assets every year. In its 2026 edition, 49% of banks and 59% of credit unions had already deployed generative AI,2 up from 16% of banks and 36% of credit unions a year earlier.3
More than seven in 10 of the 416 respondents were C-level executives.3
Generative AI Deployment Tripled at Banks in One Year
share of institutions with generative AI deployed
In a survey Computer Services, Inc. (CSI), a financial software provider, commissioned from 252 leaders at institutions of $100 million to $10 billion in October 2025,6 AI was the top technology trend for the third year running, cited by half of them, and 85% agreed that institutions adopting AI will gain a significant competitive advantage.6
A deployed share says generative AI runs somewhere in the institution, not where or how widely. A team using an AI assistant to draft member letters would answer yes. So would a bank whose online banking vendor switched a feature on in the spring.
Both can be useful, and neither says how much of your back office's work changed.
Most of It Sits Inside Software a Vendor Runs
Beyond generative AI, the kind members talk to is where credit unions are furthest ahead. Counting institutions that have invested in or deployed it, 46% of credit unions and 26% of banks have chatbots or conversational AI, against 41% and 29% for machine learning, software that learns patterns from past data, and 17% and 7% for AI agents, software that handles a task the way a trained employee would.3
Credit Unions Lead Banks on Chatbots, Machine Learning and AI Agents
share that have invested in or deployed it
View the data
That matches how community banks buy technology. In the Conference of State Bank Supervisors' 2025 survey of 268 community banks under $10 billion, 75% outsourced their core services and 66% their customer-facing technology to a third party.4
Asked which developments were promising opportunities, 46.7% picked AI for customer interactions,4 up from 31% the first time the survey asked, as Federal Reserve Governor Michael Barr pointed out.7
When the AI these banks pick sits in the customer channel, and two in three of them outsource that channel, it tends to arrive the way the mobile app did: in a vendor's release. The same banks named the responsiveness of the vendor that runs their main ledger as their second-hardest challenge to new technology, at 61.8%, up from 13% a year earlier, behind only cybersecurity.4
Community Banks Rank Their Core Processor Second Only to Cybersecurity
share of community banks
View the data
When the AI is the processor's, its pace is the processor's too, and the cost comes next: spend rate was the third-hardest challenge, at 59.2%.4
Credit unions carry one more gap. The Government Accountability Office reported in 2025 that the National Credit Union Administration had no authority to examine technology service providers, even as credit unions relied on them more for AI-driven services.8
That puts more of the weight of checking a vendor's AI on the credit union's own vendor oversight.
Fewer Than One Community Banker in Ten Is Going Institution-Wide
American Banker's research arm asked 130 bank and credit union leaders for its 2025 Data-Driven Bank survey whether they were moving forward with adopting AI tools across the whole organization.1 8% of community bankers were, against 15% at mid-tier banks and about 51% at national banks.1
For credit unions the answer was starker: not one credit union leader said they were aggressively pursuing AI adoption,1 and most community bankers and credit union leaders were exploring or running small pilots.1
What Reddit says
Bank staff on what AI does for their team so far
AI spreading for financials is starting to pop up and it does an okay job on easy stuff, but is riddled with errors on anything slightly complex.
It should just be baked into the functionality of our applications so that they anticipate what we want to do without us ever having to ask the AI agent to do anything.
American Banker published this survey in July 2025, and its 8%1 and Cornerstone's 49%2 are not one series. They answer different questions: Cornerstone's asks whether generative AI is running anywhere, American Banker's whether AI is going across the institution.
| Survey | What it asked |
|---|---|
| Cornerstone Advisors, 20262 | Has the institution deployed generative AI anywhere? |
| American Banker, 20251 | Is the institution moving forward with AI adoption across the whole organization? |
| State bank supervisors, 20254 | Which developments are promising opportunities? |
| Computer Services, Inc., 20256 | Which technology trend matters most? |
Two other figures measure what a rollout across the institution leans on: data, and people who can work with it. Only 17% of banks and 24% of credit unions rate their data strategy very effective, though both shares have risen.3 In BNY's survey of community bankers, fewer than 20% see themselves as experts in data analytics.9
The supervisors' survey shows which work community banks still handle themselves. Community banks outsource their core, but workflow processing stays largely with their own staff: 46.4% run it in-house and 40.8% with a mix of staff and vendors, while 12.8% outsource it, and it is one of the two technology areas bankers were least satisfied with.4
Community Banks Outsource the Core and Keep the Workflow
share of community banks
View the data
That is the work between systems: the file checked against the application, the reconciliation break worked from a report (two records that should match and don't), the figure retyped from one screen into another. It runs across several vendors' systems at once, so no single vendor's release covers it.
It is the work agents built for one institution take on.
FORUM Credit Union's 70% Is One Team's Estimate
FORUM Credit Union in Indiana has put a number on its lending workflow. It estimates that automated underwriting lets it process up to 70% more loans than manual underwriting, without adding staff, its chief operating officer, Andy Mattingly, told America's Credit Unions in 2025.5
What another institution can take from it is how the work is split. The AI handles routine decisions, and the underwriters keep the complex cases. For loans that come through dealers, it reviews the whole document packet: it audits the application, checks the calculations and flags inconsistencies, so staff work only the exceptions.5
None of the surveys here reports a typical gain for credit unions of a similar size, so they give nothing to set the 70% against.5
None of these surveys prices the back-office work: not one holds an underwriting time or a cost per file.
Frequently asked questions
How many community banks use AI?
No survey counts community banks alone on one measure, and the two closest ask different questions. In Cornerstone's 2026 survey, 49% of banks of $250 million to $50 billion had deployed generative AI.2 In American Banker's 2025 survey, 8% of community bankers were moving forward with adoption across the whole bank.1 The first counts AI running anywhere; the second, AI going across the bank.
How many credit unions use AI?
In Cornerstone's 2026 survey, 59% of credit unions had deployed generative AI,2 up from 36% a year earlier.3 Counting investment or deployment, 46% had chatbots or conversational AI.3 In American Banker's 2025 survey, no credit union leader said they were aggressively pursuing AI adoption.1 The answer depends on which question a survey asks.
What does deployed mean in these AI surveys?
In Cornerstone's survey it is an executive's answer that the institution has deployed it; the published results do not say where or how widely. Nothing in them rules out one assistant used by one team, or a feature a vendor switched on. None of these surveys measures how much of an institution's work runs on AI, so a deployed share says nothing about depth.
Where do community banks use AI first?
The surveys point to where members meet the institution. 46.7% of community banks named AI for customer interactions a promising opportunity,4 and at credit unions chatbots lead machine learning and AI agents, while at banks machine learning edges ahead, 29% against 26%.3 In CSI's survey, 57% of leaders named cybersecurity as AI's most valuable application.6
How far ahead are the largest banks?
National bankers were more than six times as likely as community bankers to be moving forward with AI adoption across the whole bank: about 51% against 8%, with mid-tier banks at 15%, in American Banker's 2025 survey.1 For an institution of a few hundred people, one workflow at a time is a workable way in.
Is FORUM Credit Union's 70% typical?
None of the surveys cited here gives a figure to compare it with. FORUM estimates its automated underwriting processes up to 70% more loans than manual underwriting, with underwriters keeping the complex cases.5 It is one institution's own estimate, and none of the surveys cited here reports a typical gain for credit unions of similar size. Measure your own workflow before planning on any number.
What rules apply to AI credit denials?
Where the Equal Credit Opportunity Act applies, Regulation B requires specific principal reasons for an adverse action, such as a declined application, and saying an applicant missed your internal standards is not enough.10 The Consumer Financial Protection Bureau withdrew its 2022 circular on complex algorithms and its 2023 circular on adverse action reasons in May 2025.11 The regulation still stands; confirm with your counsel.
Does AI inside vendor software count as using AI?
In the surveys, yes: a feature a vendor switched on counts the same as a workflow the institution built. 75% of the community banks state supervisors surveyed outsource their core services to a third party.4 For credit unions there is a further gap: in 2025 the Government Accountability Office found the National Credit Union Administration could not examine technology service providers, even as credit unions relied on them more for AI-driven services.8
Why is AI adoption slower across the whole institution?
The surveys do not ask why, but they show the weak spots such a rollout leans on. Only 17% of banks and 24% of credit unions rate their data strategy very effective,3 and fewer than 20% of community bankers see themselves as data analytics experts.9 Community banks also named core processor responsiveness a challenge to new technology at 61.8%.4
Are credit unions ahead of banks on AI?
On deployment, yes. In Cornerstone's 2026 survey, 59% of credit unions had deployed generative AI against 49% of banks,2 and credit unions led on chatbots, machine learning and AI agents.3 In CSI's survey, though, banks were nearly twice as likely to see AI's potential for back-office efficiency, 38% against 21%.6
What do community banks find hardest about new technology?
Cybersecurity first: 65.9% of the community banks in the state supervisors' 2025 survey named it.4 Then core processor responsiveness at 61.8%, spend rate at 59.2%, and attracting and keeping technology staff at 40.8%.4 The same banks saw AI as a promising opportunity but named implementation costs a significant barrier.4
What should a community bank measure first?
How many pieces of its own work run on AI from start to finish, and for each one, the hours each took before and after, with the mistakes caught along the way. That count says more about your institution than any adoption share, and none of the surveys measures it.
Sources
- American Banker: Banks, credit unions turn to AI for customer service boost: Report (2025)
- Cornerstone Advisors: What's Going On in Banking Research Finds AI, Crypto, and Fraud Are Top of Mind for Banking Executives in 2026 (2026)
- CU Times: CU Leaders Remain Optimistic Despite Unmet Economic Expectations: Cornerstone (2026)
- Conference of State Bank Supervisors: 2025 CSBS Annual Survey of Community Banks (2025)
- America's Credit Unions: Artificial intelligence helps one credit union boost loan processing volume by 70% (2025)
- CSI: CSI 2026 Banking Priorities: AI Emerges as Both Top Opportunity and Threat for Financial Institutions (2026)
- Federal Reserve Board: Community Banking: The Cornerstone of Building Communities, speech by Governor Michael S. Barr (2025)
- U.S. Government Accountability Office: Artificial Intelligence: Use and Oversight in Financial Services (GAO-25-107197) (2025)
- Banking Dive: BNY offers cyber, AI training to community bankers (2025)
- eCFR: 12 CFR 1002.9, Notifications (Regulation B) (2026)
- Federal Register: Interpretive Rules, Policy Statements, and Advisory Opinions; Withdrawal (Consumer Financial Protection Bureau) (2025)
