Roiva AI in Practice Index: US Regional Banks 2026
Banks with $10 billion or more in assets and $50 million to $2 billion in revenue
Which US regional banks' public disclosures show the most AI running, measured and accountable? Five leaders, A to Z.
Results as of November 12, 2026
We read what 62 US banks with $10 billion or more in total assets and calendar-2025 bank revenue of $50 million to $2 billion (in practice, about $10 billion to $50 billion in assets) have disclosed since January 2024. That covered SEC filings, earnings calls, press releases, investor presentations, company websites, vendor case studies and trade press. We then scored what that public evidence shows: how many distinct uses of AI each bank describes and how many are running, whether figures for its AI are on the record, whether it has put AI governance, Responsible AI practices and practices for the people AI affects on the record, and how strong and how recent its sources are.
AI in Practice Leaders, US Regional Banks 2026
Listed alphabetically. We don't publish scores or a ranking.
Each name links to a profile that lists the uses of AI, and the AI governance, Responsible AI and people practices, the bank has disclosed, with the date and source of each.
What does the index show, and what doesn't it?
The index shows AI that is running, measured and accountable, as a bank's public disclosures describe it: how many uses of AI are running, whether there are figures for them on the record, whether it has put AI governance, Responsible AI practices and practices for the people AI affects on the record, and how recent those disclosures are. It is built entirely from public evidence.
It doesn't show how well a bank's AI works, what it earns or saves, or how advanced the bank is. Some important limits:
- Public evidence only. A bank that says little in public may be doing more than one that says a lot. Not being named is not a finding that a bank uses little AI.
- Figures are claims, not audited results. Most are the banks' own; some come from a vendor's case study or the press. None is audited or compared with a control group. The index counts whether a figure for the bank's AI is on the public record, and how directly its source ties that figure to AI. It never uses the size of a figure, and it never compares figures between banks.
- Cash counts only when the bank credits it to AI. A financial figure counts as cash from AI only when the bank itself credits it to the AI initiative. Correlations, company-wide ratios mentioned beside AI (such as an efficiency ratio), and levels with no before-and-after earn nothing.
- Model risk management alone isn't AI governance. Banks describe model risk management, model validation and model inventories under long-standing supervisory guidance. That counts toward governance only where the bank names AI itself.
Who is included?
All 64 banks are US bank and thrift holding companies that file 10-Ks, with total assets of $10 billion or more and calendar-2025 bank revenue of $50 million to $2 billion. Two merged into another bank before the as-of date and are left out, so 62 are scored. One of them merged into a bank in the group, and its disclosures about AI count toward that bank, labeled as the predecessor's. All 62 are listed, A to Z, on the companies scored page.
The largest US banks, with more than $2 billion in annual revenue, aren't in this edition.
None is an AI-native company, so every bank scored is ranked.
How was the index made?
Working under the direction of Roiva staff, AI research agents (Claude, made by Anthropic) searched each bank's disclosures about its AI, how it governs AI and AI's effect on its people, and copied the exact words of every source they used. A script, and in some cases a separate checking agent, confirmed each quote on its source page. Roiva staff set the method and its weights. Before a bank is named a leader, a Roiva staff member checks every source behind it.
The full method, including the weights, the leader rule and the notes for banks, is on the method page. Results are as of November 12, 2026.
A separate Roiva study, What mid-size US banks and credit unions disclose about AI, reads a random sample of 50 banks in the same revenue band, at a wider range of asset sizes, and 50 credit unions. Its sample and counts differ from this edition's.
Is your bank in the index?
- Every bank in the study can ask for its own score and the sources behind it, free of charge. We share scores privately; we don't publish them. We confirm each request with the bank, through a contact we find independently, before we send its score.
- Evidence that isn't public never counts toward this index, in any edition. It can be assessed only in the separate Roiva Verified AI Rating (below).
- Claim your profile. A bank that isn't named a leader can ask for a public profile. It is published only if the bank opts in and its public evidence clears the bar set out on the method page.
- Add your own section. Any bank with a public profile, leader or not, can add a short company-provided section to it, under our terms. It never counts toward the index.
-
Disputes and corrections. If we've missed a public disclosure, misread a source or
got a date wrong, write to data@roiva.ai with a link to the public source.
- We reply within one week, and reach an outcome within 30 days.
- We only add or change evidence that anyone can check in public.
- When a correction changes this list or a profile, we update it and record the change on the change log.
- Remove a profile, or leave the list. A bank can ask, free, for its profile page to be taken down, and a leader can ask to come off this list. Scores don't change, and nobody moves up in its place.
Leaders who asked not to be listed: 0.
Is there a rating based on private evidence?
Yes: the Roiva Verified AI Rating is separate from this index. Any company, named in the index or not, can ask to have evidence it holds privately assessed, such as records of an AI use running or of a Responsible AI practice operating. The rating is opt-in, and published only if the company chooses. Its fee is $2,500, $7,500 or $15,000 a rating, set only by the company's annual revenue and the same for every result, paid before the review starts. Nothing in it ever changes an index result, and being rated, or not, never affects whether a company is named here. Our evidence guidelines explain what to send and how it is handled.
About Roiva
Roiva makes software that helps companies measure the return on their AI initiatives, and may sell to companies in this index.
Our commercial relationships with them are disclosed as counts, never names. As of November 13, 2026, none of the 62 banks scored is a Roiva customer, partner, supplier or investor. As of the same date, Roiva, its staff and their households hold no shares or other positions in any of them.
No company can pay for a place, a badge or a profile, and being a customer never changes a score. See our independence policy. Being named does not mean Roiva endorses a company or its products, or that the company endorses Roiva.