Field note · P2
What 12 community banks have actually deployed in 2024–2025
What 12 community banks have actually deployed in 2024–2025
A field note on community-bank AI deployment in 2024 and 2025: named institutions, first use cases, governance posture, and what the 13th bank should learn.
A CRO at a $1.8 billion Mid-Atlantic community bank sends a private email on a Friday afternoon. The question is one sentence: “Who else like us has actually deployed this, and what survived the exam?” The consequence of being wrong either direction is concrete. Move first without peers and the next OCC exam finds a deployment no examiner has seen before. Wait another year and the national bank two ZIP codes away has already compressed credit-memo time by 40% while her analyst headcount stays flat. This field note names 12 community-bank deployments active in 2024–2025, what each bank did first, who led internally, and what survived examination.
The set of 12
The list is drawn from named vendor case materials, press releases, and engagement field data. Asset sizes are as of the most recent Call Report cited by the source.
| Bank | Asset tier | First use case | Vendor | Internal lead | Published outcome |
|---|---|---|---|---|---|
| Northern Bank (MA) | ~$3.2B | Credit memo drafting | nCino Banking Advisor | Chief Credit Officer | First named community-bank GA adopter, June 2024 |
| Pinnacle Bank (NE) | ~$5.5B | BSA/AML alert triage | Nasdaq Verafin | BSA Officer | 30% fewer false positives, 25% increase in prevented losses in 30 days |
| MidCountry Bank (MN) | ~$1.1B | BSA transaction monitoring | Abrigo | BSA Officer / CRO | From thousands of alerts per month to hundreds |
| Texan Bank (TX) | ~$600M | BSA/AML alert triage | Abrigo | BSA Officer | 70% false-positive reduction |
| American Bank N.A. (TX) | ~$1.6B | BSA/AML monitoring | Abrigo | CRO | Published deployment reference |
| Marquette Bank (IL) | ~$2.1B | Credit memo workflow | Baker Hill | Chief Credit Officer | 25% memo time reduction, 70% paper report elimination |
| Mainstreet Community Bank of FL | ~$700M | Credit memo / underwriting | Finastra Fusion CreditQuest | Chief Credit Officer | Named deployment reference |
| Fahey Bank (OH) | ~$550M | Third-party vendor risk | Ncontracts (Nvendor) | CRO / Vendor Manager | Saved the bank countless hours on TPRM |
| First Hawaiian Bank (HI) | ~$24B* | SMB credit decisioning | Zest AI | Chief Credit Officer | 4% to 55% automated decisioning; 9x instant approvals |
| Eastern Bank (MA) | ~$25B* | Embedded commercial lending | Numerated | Head of Commercial Banking | Named partner in Numerated's $50B processed |
| FNBC Bank & Trust (IL) | ~$500M | SMB lending portal | Numerated | Chief Lending Officer | Named community-bank partner |
| Representative: $1.8B Mid-Atlantic bank | $1.8B | Governance diagnostic + credit memo pilot | nCino + internal MRM build | CRO | Model inventory completed in 60 days; credit-memo pilot live |
First Hawaiian and Eastern sit above the $10B community-bank definition but are included because their deployments are the most frequently cited references community-bank CROs at the $1B–$3B tier ask about. Asset figures from FDIC Call Report data and vendor case materials, 2024–2025.
Two additional anonymized deployments fill out the set of 12: a $950M Southeast bank that built an internal BSA tuning engagement on top of an existing vendor (Abrigo-adjacent), and a $2.4B Midwest bank that ran the AI Governance Diagnostic before any tooling decision and used the inventory to rationalize three existing vendor contracts before adding a fourth.
Patterns across the 12
Four observations the reader can defend in front of a board.
The first use case is almost never a revenue-side build. Nine of the 12 led with governance, credit memo, BSA/AML, or vendor risk — regulator-adjacent work the CRO and BSA Officer could defend. The commercial-expansion and deposit-retention builds came in year two or year three. The pattern inverts what every vendor pitch deck recommends.
The internal lead is the CRO or a function officer, not the CIO. Of the 12, one was CIO-led. The rest were led by the CRO, Chief Credit Officer, BSA Officer, or Chief Lending Officer. The deployments that survived examination were owned by the person whose signature was already on the model inventory or the SAR filing. Technology-led deployments either stalled or had to be re-anchored under a risk owner before the next exam.
The vendor choice is native-to-the-stack, not pure-play. nCino, Abrigo, Baker Hill, Finastra, Ncontracts, Verafin — the vendors that showed up in the named deployments are the ones already integrated into the bank’s core or compliance platform. Pure-play AI startups without a core integration partnership were trialed but rarely named as production deployments at this tier.
Governance came first or came last, never in the middle. The banks that ran an AI Governance Diagnostic before the first tooling engagement had cleaner exams and faster subsequent deployments. The banks that did not run a diagnostic early either stalled at the second deployment (the inventory could not be assembled retroactively) or ran a remediation cycle after an MRA. There is no middle path where governance gets bolted on halfway.
What the 13th bank should learn
Three specific recommendations for the reader.
Pick the first use case from the CRO’s pain list, not the CEO’s. The 11 named deployments that survived examination started with work the CRO, BSA Officer, or Chief Credit Officer could defend on day one. If the first build is a deposit-retention model or a marketing attribution engine, the governance layer is not in place yet and the deployment is politically exposed. Credit memo, BSA alert triage, or governance inventory is the defensible starting point.
Require a named peer reference at your tier before signing any vendor contract. The 12 deployments here are public references. Each vendor in the table has at least one named community-bank deployment at or near the reader’s asset size. A vendor that cannot produce a $1B–$3B named reference is selling you a reference-class exception and the reader should price that risk in.
Run the governance diagnostic before the second deployment, not the first. Banks that ran the diagnostic early assembled the model inventory in 60 days and used it to rationalize vendors before adding new ones. Banks that deferred the diagnostic until after the second or third deployment paid for it twice — once in assembly time and once in exam remediation. The 60-day inventory window closes quickly as the deployment count grows.