Every leadership team has had the meeting. Someone says, "We should look into AI." Heads nod. The agenda moves on. Three months later, someone says it again.
Nothing was decided, so nothing happened. The fix is not a better strategy document. It is a better meeting. This AI strategy meeting agenda gives you 90 minutes, five decisions, and one outcome: a named person, a named process, and a date.
If you want the longer-term plan, we cover that in how to build an AI strategy that connects to results. Think of this meeting as the one that starts it. The strategy is the document. The meeting is where someone finally agrees to own it.
Why the AI conversation keeps stalling
Three patterns repeat in leadership teams at companies with $5M–$500M in revenue.
The topic is too big. "AI" is not an agenda item. It covers everything from a chatbot on your website to rewriting how you quote jobs. A group cannot decide something that large, so it defers.
Nobody knows the starting point. Leaders discuss what AI could do without knowing what their own people already do with it. The conversation floats above the real company.
No one is on the hook. A meeting that ends with "let's all think about it" has assigned the work to everyone, which means no one.
The agenda below answers each pattern. It narrows the topic to one process, starts from facts, and ends with an owner.
Before the meeting: two things to do this week
Do these before you put 90 minutes on the calendar.
- Invite the right five to eight people. The owner or CEO, plus the leaders who run operations, finance, sales or service, and technology (or whoever handles IT). Add one person who works closest to the customer. Skip observers.
- Send one question ahead. Ask each person: "Where does your team use AI tools today, including on personal accounts, and what is the one task that eats the most hours?" Have them reply in writing, by name if they are willing.
That second step is the single most useful thing you can do today. The answers become your raw material for the first hour.
The 90-minute AI strategy meeting agenda
Minutes 0–10: Set the rules
State the purpose in one sentence: "By the end of this meeting we will have chosen one process, one owner, and one review date." State what the meeting is not. It is not a vendor demo, a tool comparison, or a debate about whether AI is good or bad.
Agree on one ground rule: every item that comes up is either decided, assigned, or parked on a written list. Nothing is left floating.
Minutes 10–30: What we already use
Start with a shadow AI inventory. Shadow AI is any AI tool your people use that the company has not approved or does not know about: a free chatbot to draft customer emails, a browser extension that summarizes documents, an app that transcribes calls.
Use the written answers you collected. Put them on a shared screen and sort them into three columns:
- Known and approved. Tools the company chose.
- Known but unapproved. Tools you found out about and have not ruled on.
- Unknown. Gaps in the list. Ask who else might be using something and has not said.
Do not use this time to scold anyone. People who adopted tools on their own were usually trying to do good work faster. If they fear punishment, the next inventory will be shorter and less honest. Our piece on what employees fear about AI explains why that matters.
One decision to leave this block with: which unapproved tools stop today, which are allowed for now, and who writes the rules. If you do not yet have a written policy, writing an AI use policy is the follow-up.
Minutes 30–55: The one process that hurts most
Now pick one. Not three, not a "top five." One.
Ask each leader to name the process that costs the most time, causes the most errors, or frustrates customers the most. Write each on a whiteboard. Then test each candidate with four questions:
- Is it repeated often enough to matter?
- Can someone describe the steps on one page today?
- Does a mistake cost you something you can measure, such as hours, dollars, or customer complaints?
- Does a person still review the output?
A process that fails the second question is not ready. If nobody can write down how the work gets done, AI will automate the confusion. That is a finding in itself, and documentation before AI covers how to fix it first.
Vote, then decide. The room should end this block with one process written at the top of a page, plus a plain-language statement of what "better" would mean. Leave the number blank if you do not have a baseline. A made-up target is worse than none. Your first assignment may well be to measure how long the work takes now.
Minutes 55–70: What we will not automate
Most AI meetings skip this block. It may be the most important one.
Name the work you will keep in human hands, and say why. Candidates include:
- Conversations where a customer is grieving, angry, or making a hard decision.
- Hiring, firing, and performance decisions about named people.
- Pastoral or counseling-adjacent moments, if your company serves people in that way.
- Anything where being wrong creates legal exposure, until your review process is proven.
- Work your team takes pride in and would be diminished by giving away.
For a Christian-led company, this block is where stewardship becomes concrete. Stewardship is not only about saving cost. It includes how people are treated and what their work means to them. Our article on AI and the dignity of work goes deeper. For a secular company, the same list protects trust with customers and staff.
Write the list. Share it with the team after the meeting. A short "we will not" list does more for adoption than any amount of enthusiasm.
Minutes 70–85: Who owns the next step
Pick one person. Not a committee, and not "IT and operations together."
The owner needs three things: enough authority to ask other departments for time, enough credibility that people will answer, and a direct line to you. Often that is the operations leader. Sometimes it is the CFO. It is rarely the most junior person who "knows tech."
Then define the owner's first deliverable. Keep it small enough to finish in 30 days. Examples:
- A one-page map of how the chosen process works today, with the time each step takes.
- A written list of the data the process touches and where it lives.
- A short list of the questions the team needs answered before any tool is chosen.
If the owner cannot do this alongside their current job, say so now and decide what comes off their plate.
Minutes 85–90: The date we review
Set a date on the calendar before anyone leaves the room. Thirty to sixty days is about right for a first review. Longer, and the topic goes cold.
At that review the owner reports three things: what they learned, what the baseline is, and whether the group should move forward, change direction, or stop. Stopping is a legitimate outcome. A review that can only say yes is not a review.
Common mistakes with an AI leadership meeting
Starting with tools. If the first slide is a product, you will spend the meeting comparing features instead of deciding where the pain is.
Letting the loudest enthusiast run it. The person most excited about AI is not always the person who should own the work. Match the owner to the process.
Skipping the inventory. If you do not know what your people already use, you cannot write a sensible policy, and you may be carrying data risk you have not seen. See AI security for business for what to check.
Treating the meeting as the finish line. Ninety minutes gets you a decision. It does not get you a result. The result comes from the work the owner does in the next 30 to 60 days.
Hiding the mess. Some of what you find will be uncomfortable: undocumented processes, tools nobody approved, a leader who is skeptical. Put it on the table. A clear picture of a messy situation beats a tidy picture of an imaginary one.
What to do with the notes
Within 24 hours, send the team a one-page summary: the tools found, the process chosen, the "we will not automate" list, the owner, the first deliverable, and the review date. That page is your decision record.
It is also what makes the next conversation productive. Outside input is far more useful when it starts from a record of what you decided and why, rather than from a blank page.
Your next step: bring the notes to a Discovery Call
Run the meeting. Then book a Discovery Call and bring your notes. It is 30 minutes, no cost, no obligation. We will read what you decided, tell you plainly whether the process you chose is a sound first step, and say so if the answer is "not yet" or "smaller." We make recommendations and referrals, and we may be compensated for some of them, but there is no tool you need to land on for us to tell you the truth about it.
If you would like a quick read on where you stand before you schedule the meeting, the AI Capability Score takes about five minutes. It is a useful starting point for the first conversation, not a substitute for it.