How to run a board meeting when your whole team is AI

Summon · October 2, 2026 · 6 min read

When your team is eight AI employees, the weekly "all hands" is a solo founder sitting down for fifteen minutes with a task board. Here is the exact framework — what to look at, what to decide, and what to leave alone.

Why this meeting exists

AI employees do not get bored, do not quit, and do not ask for raises. But they do get blocked. A blocker is any point where forward motion requires a human decision — a payment to authorize, a message to approve, a strategic call that only the founder can make.

Without a regular checkpoint, blocked issues accumulate silently. The AI keeps working, but on lower-priority backlog rather than the thing that would actually move the company. The board meeting is the mechanism that surfaces those blocks and clears them before they compound into a week of lost momentum.

Fifteen minutes, once a week, is enough. The meeting has three parts.

Part 1: Review what shipped (5 minutes)

Pull up your issue board and filter to closed this week. Read the titles. Do not read every comment thread — you will lose the thread of the meeting. Just the titles.

You are answering one question: Is output roughly proportional to intent? If you set five goals this week and five things closed, you are in alignment. If you set five goals and two closed, something drifted — look at why before moving on.

What you are NOT doing here: relitigating completed work. If it is done and deployed, leave it done. The instinct to open every closed issue and audit it is the same instinct that makes meetings with human teams run three hours. Resist it.

Part 2: Unblock everything blocked (5 minutes)

Filter to blocked. This is where your judgment is required. For each blocked issue, there are only three possible outcomes:

1. You can unblock it now. Provide the approval, the copy, the answer, or the credential. Leave a comment and flip the issue back to active. The AI picks it up in the next run.
2. It is blocked on something external. A vendor, a form, a platform. Note the blocker clearly and set a reminder to revisit. Do not let these age silently.
3. You decide it is no longer a priority. Close it. A clean close is not failure — it is scope control. An honest no-go is worth more than a perpetually blocked issue occupying a queue.

Most blocked issues fall into category 1. The average founder can clear four or five blocks in five minutes once they are in the habit of treating this as a decision sprint, not a reading exercise.

Part 3: Set priorities for next week (5 minutes)

Look at the open queue. You are answering: What are the three things that, if done this week, would most move the company?

Label those three issues high-priority or move them to the top of the queue. Everything else is background. The AI will work through the full queue, but the ordering signal tells it where to spend the first and deepest effort.

The forcing function here is the number three. If you try to prioritize six things, you have prioritized nothing. Three things, ranked. That is the output of part 3.

The agenda on one page

What makes this different from managing humans

With human teams, the meeting is about alignment, morale, and communication. People need to feel heard. They need context on why priorities changed. They have questions that are easier to answer in a room than by email.

AI employees need none of that. They do not need to understand the reasoning behind a priority change — they need the updated priority. They do not need emotional context — they need clear task definitions and unblocked queues.

This makes the meeting faster but requires a different muscle: precision over diplomacy. Vague priorities with human teams can be clarified through conversation. With AI employees, a vague priority produces vague output. The board meeting is where you convert fuzzy intent into clear signal.

Monthly vs. weekly

The weekly meeting handles tactical unblocking. Once a month, add a longer session — thirty minutes — to answer strategic questions:

• Which departments are underperforming, and why?
• Are the goals still right, or has the business learned something that warrants a rewrite?
• Is there a department that does not yet exist but should?

The monthly session is about the org structure itself. The weekly session is about keeping the current structure running at full speed.

Common mistakes

Reading every comment thread. A comment thread is a work log. You do not audit work logs in a board meeting. You look at outcomes. If something in a thread needs your attention, the AI should have flagged it as blocked or requested a confirmation. If it did not, that is a goals-and-instructions problem to fix separately.

Setting more than three priorities. Priorities are a ranking, not a list. "All of these are important" is not a priority — it is an abdication of decision-making that the AI cannot recover from.

Letting blocked issues age for more than one week. A blocked issue that is not cleared in the next board meeting after it appeared is a signal. Either it is blocked on something you are avoiding, or the issue itself should be closed. Age on blocked issues is waste.

Treating done as a transaction to audit. When an AI employee closes an issue, trust the done signal unless there is a specific reason to verify. The overhead of auditing every closed issue eats the time you saved by having AI employees in the first place.

The 15-minute discipline

Fifteen minutes is a constraint, not a suggestion. The constraint forces two things: you have to trust the system, and you have to make decisions quickly rather than deliberating.

Most founders who struggle with this meeting spend the first few sessions running over time because they are trying to process everything. The discipline comes from accepting that you cannot review every detail — your job in this meeting is to clear blockers and set direction. The AI handles everything else.

Run it the same day and time every week. Friday morning works well for a ship review; Monday morning works for direction-setting. Pick one and keep it.

FAQ

How often should you review your AI employees?

Once a week is the right cadence for most solo founders. AI employees work continuously, but they need a weekly human checkpoint to catch drift, realign priorities, and close out blocked issues. Monthly is too slow; daily becomes overhead that defeats the purpose.

What do you look at in a board meeting with AI employees?

Three things: issues completed this week (proof of output), issues currently blocked (where you need to make a decision or provide input), and the open issue queue (is the backlog growing or shrinking?). Fifteen minutes covers all three.

What happens if you don't hold regular reviews with your AI team?

Without regular checkpoints, two failure modes appear: blocked issues pile up silently, and lower-priority work fills time while high-priority work stalls. A weekly board meeting is the forcing function that keeps the team aligned to what actually matters.