Twelve weeks and change remain in 2026, and at least two of them disappear into holidays. Whatever your credit union did with AI this year, four items decide whether the work compounds in 2027 or resets in January. Each one is a documentation and decision task, not a new project, which is exactly why Q4 is the right window: no procurement, no integration, just closing loops your team already opened.

1. Document the verdict on every pilot

If your credit union ran a pilot this year, somebody owes it a written verdict: scale it, fix it, or kill it. In practice the verdict that gets written down is usually none of them; the pilot keeps running, the license keeps billing, and by March nobody remembers what the success metric was supposed to be. Pull up the measurement plan from the 90-day pilot structure and record the decision with the numbers behind it, even when the numbers are embarrassing. A kill memo with real figures is worth more next year than a pilot nobody will defend. If the answer was scale, the program playbook covers assigning the owner and picking workflow number two.

2. Get the AI budget category into the 2027 budget

Budget season is closing at most institutions. If AI spending is still scattered across software, consulting, and miscellaneous lines, 2027 will be another year of invisible spend and surprise totals. The case for a dedicated 2027 AI budget category is one conversation with your CFO, and budget calendars do not reopen in February because the category made sense. If your budget already locked, ask for the tracking category anyway; visibility does not need a new appropriation.

3. Review every AI-touching contract that renews in 2027

List the vendor contracts renewing next year and flag the ones where AI features arrived mid-term, because several of your vendors shipped AI into your stack this year without asking. Check each against the terms in the vendor contract guide: data rights, model training on member data, audit artifacts, exit provisions. Most auto-renew clauses want 60 or 90 days’ notice, so a contract review that happens in December is a reading exercise, not a negotiation.

4. Bring the exam file current

Examiners are asking AI questions now, and the answers live in documents: the tool inventory, vendor due diligence records, the monitoring evidence for anything in production. Run the exam readiness checklist against what is actually in the file, and right-size the gaps using the model risk framework rather than importing big-bank process. An hour of gap-checking in October beats a scramble the week the exam letter lands.

The one-meeting version

Put all four on the agenda of one ops meeting: pilot verdicts documented, budget category submitted, renewal list reviewed, exam file checked. None of them need a vendor, and all four are the difference between an AI program and a pile of experiments. More screening material is in the back-office automation pillar.

If you want a second set of eyes on the gaps before year end, Advisor Labs runs a 45-minute AI readiness audit for credit unions: book a conversation.

Get this roundup in your inbox every Thursday: subscribe to the AiForCU newsletter.