An explainer for community banks and credit unions — what “owned intelligence” is, where it helps, and how to evaluate AI and OI.
Engagements begin 2027 · shown by invitation
Every institution is being offered artificial intelligence right now. Almost all of it arrives the same way — as a service, from somewhere else, priced by the month. There is a second method, and the difference matters more in a supervised institution than almost anywhere else.
Claude, ChatGPT, Gemini and the rest work by prediction. A program read an enormous amount of text and learned the patterns in it (training). When you ask it something, it does not look anything up — it produces one word, then guesses the most plausible next word, then the next, until an answer has formed.
There is no library inside. Nothing is consulted. That is why these machines are fluent, fast, and frequently right — and also why they can be smoothly, confidently wrong.
For a machine built this way, inventing an answer feels exactly the same as knowing one. It cannot tell the difference, and neither can the reader, because both arrive in the same confident voice.
It also lives somewhere else. It runs in a distant company’s data center, and whatever is typed into it travels there. The arrangement is a dependency: it works while you pay, while you are connected, and while that company continues to offer it. For a great many uses this works well. For work a bank examiner will read, or that touches a member’s private information, it is not a great fit.
The same underlying technology, in a different configuration. Three changes, and they are the whole idea.
It runs on a closed machine with no line out to a distant company. Nothing you hand it is typed into a service; no outside company sees a borrower’s file, sets the terms, or can switch it off. Who holds that machine is the subject of Part Three.
It works from a shelf: your documents, and the rules that govern them. Its job is to find, compare, and quote — not to recall. Every figure points back to the page it came from.
What it produces is a draft. Where a decision belongs to a person, it stops and waits. Someone then reviews the work item by item and signs, or doesn’t.
Put plainly: AI is a brilliant, confident consultant who has read everything, cites no source, keeps none of your files, and rarely says “I don’t know.” OI is a very fast assistant who works strictly from your documents, shows you the page behind every number, and leaves the report on your desk for your review and signature.
Both are useful. Only one of them can be verified. Verifiable is not verified — the citation makes checking take seconds, but a person still checks.
| AI — as usually sold | OI — owned | |
|---|---|---|
| Where it runs | A distant company’s data center. Your questions travel there. | A closed machine with no line out — ours during an engagement, yours when you want one of your own. |
| How it answers | By prediction, from patterns — or from documents held on someone else’s computers. | By finding and quoting your own documents and rules. |
| When it doesn’t know | It often produces something plausible anyway, in the same confident voice. | It stops, and says so, because it was built to. |
| What it costs | A meter that never stops — every answer generated again from scratch. | A fixed fee, agreed in advance. The price does not move with how much you use it. |
| Who sees your information | The company running it, under terms it can revise, for as long as it keeps it. | During an engagement, us — under written terms, for the length of the work, nothing retained after delivery. No outside company, ever. |
| If the company disappears | The service stops. | You keep the work, the shelf, and the record. So do the paper files, as always. |
| Who is accountable | Unclear — no one signed anything. | A named person at the institution, per item, on the record. |
Regulation does not scale down. A credit union with nine employees answers to substantially the same rulebook as one with nine hundred, and prepares for substantially the same examination. The work does not shrink with the balance sheet — it just lands on fewer desks, usually belonging to people who are also doing three other jobs.
That is the gap. Not a lack of technology. A lack of hours, in an institution that cannot solve the problem by hiring a compliance department.
The work begins by finding where the hours actually go. What follows is what that search turns up — not a menu of features, but the places the friction was.
1040s, K-1s, Schedule E, the entity returns. Form, year and line number are the same at every institution in the country, so this is read once and read the same way every time. What is not the same — add-backs, K-1 treatment, distributions against retained earnings, contingent liability haircuts — is your method, signed by you, applied the same way to every file.
Every reporting requirement, financial covenant, insurance and tax obligation in every agreement — read out, dated, and matched against what is actually in the file. Most institutions cannot answer this today, and it is where credit-administration findings come from.
Related entities, pass-through structures, cross-guarantees, and debt held at other institutions. Assembled from documents your covenants already require and your files already hold, read as a set rather than filed one at a time.
The first-day letter is mostly questions about records. Those records are produced by the work itself — who decided, on what basis, against which edition of the rules and which edition of your method, and when.
The institution’s memory lives in files rather than in whoever has been there longest — which matters most in a shop small enough that one retirement is a crisis.
Commercial and industrial credit runs on a faster clock than anything else on the balance sheet, and the discipline it demands is relentless rather than difficult. Borrowing base certificates arrive monthly and are recalculated by hand. Compliance certificates fall due quarterly. Interim financials, annual statements, tax returns, insurance renewals, UCC continuations, field exams and appraisals all run on separate schedules that no single spreadsheet tracks well.
Eligibility rules read from the agreement, advance rates and sublimits applied, ineligibles and concentration caps computed, availability compared against the outstanding balance — with every figure pointing back to the certificate it came from.
Fixed charge coverage, leverage, tangible net worth, minimum liquidity — calculated from the statements as they arrive, against the definitions in that borrower’s own agreement rather than a generic formula. The arithmetic is stated; naming a breach remains yours.
Monthly, quarterly and annual obligations tracked per credit, aged when late, with the request letters drafted. A C&I book generates more reporting events in a quarter than a CRE book does in a year, and that is exactly why it goes uncollected.
Commercial real estate runs on an annual rhythm, which makes it easier to defer and easier to lose track of entirely.
Every property, every other lender, the guarantor’s global cash flow. Collected annually by most institutions under a covenant they already wrote; assembled as a set by almost none.
Loans written at the 2021–22 trough are repricing now. Coverage recomputed at reset, property by property and across the portfolio, under your own signed method — before the examiner asks.
Flood certification, insurance, title, appraisal, entity papers — the required set derived from the loan’s own characteristics, with staleness tracked. Nothing discovered at the exam that could have been caught at closing.
It runs on the documents you already have, and hands the result to a person.
Most of what is sold in this field arrives as a system your staff must learn, operate, and answer for from the first day. That is the wrong order. The work comes first; the machinery follows the institutions that want it.
A fixed-scope engagement, four to six weeks. We read your agreements and files on our own closed machines, build the shelf with you, and deliver the obligation inventory, the exception report with drafted request letters, the assembled borrower positions, and the reset study. Every figure cites its page or its cell. Your officer reviews and signs. Nothing of yours is retained after delivery, and nothing is ever typed into a rented service.
A census is a discovery; the obligations are perpetual. An annual arrangement keeps the inventory current, refreshes coverage as documents arrive, and delivers a quarterly exception report and an annual reset study. Same discipline, same citations, and still no system for your staff to run.
Some institutions will want this running on their own premises, on their own schedule, operated by their own people. That is the device this explainer describes, and it will be built for the institutions that ask — after the shelf is built, the method is proven on their own files, and there is something worth installing. Not before.
The commitments in Part Five hold at every stage. What changes is who holds the machine, not what it is allowed to do.
Named in advance, before any demonstration: credit decisions — approving, pricing, declining. Risk ratings and grade migration. Whether a covenant breach is an Event of Default. Suspicious-activity determinations. Fair-lending judgments. Adverse action. Anything where the wrong word is an enforcement matter rather than a correction. The machine drafts; it does not decide.
The work may state that a tested covenant computes as not satisfied under your signed method, with citations. Naming that as a default is a credit decision and stays with your credit officer.
A tool that claims to do all of this is worth questioning. A stated ceiling, in writing before you buy, is what your examiners will look for.
Stated before anything is sold, because these are the parts that would be easiest to quietly abandon later, and because an institution should be able to hold a vendor to something in writing. They hold whether the machine sits on our desk or yours.
No member or borrower record is typed into a rented service, transits a third party, or trains anyone’s system. During an engagement your documents are held on closed machines we control, under written terms, for the length of the work, and returned or destroyed at delivery. When the machine is yours, they never leave your building at all.
What the machine consults — which regulations, which editions, which of your own policies, and how coverage is calculated — is assembled together and signed by your institution before the first pass runs. Nothing goes on the shelf without your signature.
A defined set of documents, for the length of one pass. It holds no credential to your core, your email, or any member system. There is no share link and no copy anywhere else, and every export is logged.
Regulations carry effective dates; so does your calculation policy. Every piece of work cites the editions it ran against, and a change is a new dated version you review and adopt — never applied retroactively, never pushed to you overnight.
Human review is not a checkbox added at the end. Work is sized to what attention can actually cover, and material figures are confirmed one at a time, so that rubber-stamping is harder than reading.
Checking happens against the published rule and the original document — by plain machinery, or by a person. Nothing is ever asked to grade its own work, and every number in a ratio is arithmetic, not a prediction.
Your files are not our laboratory. Development happens on invented data.
We own that the process was followed, that every item is cited to its source, and that the record proves it. What the rules require of your institution is yours to decide, with your counsel. We assemble, we flag gaps, and we never pretend to be your compliance advisor.
The shelf, the record, the citations, and the deliverables stay with you and stand on their own. Where a device is installed, the documentation stays with you and the software is escrowed for release. The last question below is one we ask you to put to any vendor — this is our answer, in writing.
Ask the vendor who brings artificial intelligence to your financial institution:
If the answer involves anyone else’s computers, ask what happens to it there, who can read it, how long it stays, and under what terms — in writing, before the work starts.
If the answer is anything other than “it stops and says so,” ask what it does instead, and how you would know.
Every figure should point somewhere a person can check in seconds. A number without a source cannot be verified.
Ask to see the calculation policy in writing, with a date and a signature. If coverage comes out at 0.86 rather than 0.94, the answer should be a method your institution adopted — not a model’s judgment.
Ask to see what the reviewer actually sees. If approving a hundred items takes one click, the review is nominal.
If the price rises with use, you have rented a service rather than bought a tool — which may be fine, as long as everyone knows which one it is.
A small vendor is not a disqualification, but the honest ones have an answer that does not require them to survive. Ours is in Part Five.
smallbanc is taking a small number of engagements beginning in 2027, one region at a time.
Every engagement is a field test. The record the work produces is built to stand in front of an examiner, and the results will be published as they accumulate.
Write to us