Leadership ┬╖ 4 minute read
How to Run an AI Town Hall
An AI town hall works when it is specific: which processes change and when, what agents will do and what people will do instead, how measures change, and what happens to freed capacity, including saying when a decision has not been made. Take unfiltered questions, answer the headcount question directly, and follow through in writing.
An AI town hall is where leadership credibility on AI is won or lost. Employees arrive with a specific question they may not ask aloud, having read the same headlines as everyone else. A session of generalities confirms their fears; a session of specifics, including uncomfortable ones, buys the cooperation the program depends on. This guide covers what to say, what will be asked, and what to do afterward.
What should the prepared remarks contain?
Five things, in this order, and all specific:
- What is changing: the named processes and the timeline. Not "we are exploring AI across the business."
- What the agents will do, and what people will do instead, by role.
- How roles and measures change: titles, expectations, performance measures, and when.
- What happens to freed capacity: reinvested where, redeployed how, or reduced by what process. If undecided, say so with a date.
- How to report problems with the agents, and that doing so is expected and safe.
The how to communicate AI changes to employees guide covers the wider communication plan this session sits inside.
How should the headcount question be handled?
Directly, and in the prepared remarks rather than waiting to be asked. Everyone in the room is thinking about it; addressing it first signals that leadership is not avoiding it, and the rest of the session is then heard differently.
The three honest answers are reinvest, redeploy, or reduce, and most companies use a mix by function. If the decision has not been made, say precisely that, explain what it depends on, and commit to a date. What destroys credibility is a vague reassurance that later proves false, and employees have heard enough of them to recognize the pattern. The how to think about AI and headcount guide covers making the decision.
What questions will be asked?
| Question | What is really being asked | How to answer |
|---|---|---|
| Will this cost jobs? | Am I safe? | The specific answer for their function, or the date it will be decided |
| What happens to my role? | Will my work still matter? | The redesigned role, concretely |
| What if the AI gets it wrong? | Will I be blamed? | The reporting process, and that reporting is expected |
| Is this being forced on us? | Do we have a say? | Who is involved in design, and how to join |
| Why are we doing this? | Is it cost cutting dressed up? | The thesis, honestly, including cost if that is part of it |
| Will I have to learn to code? | Am I going to be left behind? | The actual skills: supervision, exceptions, specification |
| Who decided this? | Is anyone accountable? | Name the accountable executive and the owners |
Anticipate these in the prepared remarks and the question session becomes a conversation rather than an interrogation.
Should questions be pre-screened?
No. Pre-screened questions confirm exactly the suspicion the session exists to address, and employees notice immediately. Take live questions plus anonymous submissions, answer what you can, and commit to written answers for what you cannot. One hard question answered honestly does more for trust than an hour of smooth presentation.
Who should speak?
The executive accountable for the program, and the leaders of the functions affected. Employees want to hear from the people who decide about their work. A technology leader can explain the systems briefly, but a town hall run by the technology function is heard as a project update, not as a commitment about people's jobs.
Brief managers beforehand, in detail. A manager who learns the plan alongside their team cannot support them afterward, and the conversations that matter most happen in teams the following week, not in the town hall.
What happens afterward?
- Written answers to every question asked, published where everyone can see them, including the ones that were awkward.
- Answers to open items by the dates promised. Missing a promised date on the headcount question is particularly damaging.
- Manager conversations in affected teams.
- A follow-up session that reports what actually happened against what was said, including where the plan changed.
Town halls without follow-through are worse than no town hall, because they establish that announcements are not commitments. The how to build an AI-first culture guide covers the norms that follow-through builds.
What should executives ask themselves?
- Could an employee leave this session and name the processes changing and when?
- Did I answer the headcount question before being asked?
- Did I take a question I would rather not have?
- Are managers briefed well enough to handle next week's conversations?
- What did I promise, and when is it due?
How can FISTA Solutions help?
FISTA Solutions works with executive and HR teams through its AI enablement practice on process redesign that involves the people doing the work and on the communication that follows, and builds AI agents with the reporting channels and exception handoffs that make the promises in a town hall real. Since 2017, FISTA has delivered 150+ projects for 50+ companies across 12+ countries.
To prepare the specifics before your next all-hands, talk to FISTA on WhatsApp, or read how to redesign jobs around AI agents.
Share-ready article cover
Download the generated social format.
Clear answers
Questions raised by this field note.
Straightforward guidance for evaluating scope, fit, and the next step.
01What should leaders say in an AI town hall?
Which processes are changing and when, what the agents will do, what people will do instead, how roles and measures change, what happens to the capacity that is freed, what is not yet decided and when it will be, and how to report problems with the agents. Specific, not general.
02How do you answer the headcount question?
Directly, in the prepared remarks rather than waiting to be asked. If capacity is being reinvested, say where. If roles will reduce, say how, when, and with what support. If the decision is not made, say so and give a date. Evasion on this question discredits everything else said.
03Should AI town hall questions be pre-screened?
No. Pre-screening confirms exactly the suspicion the town hall exists to address. Take live questions, including anonymous submissions, answer what you can, and commit to written answers for what you cannot. A hard question answered honestly does more good than a smooth session.
04Who should speak at an AI town hall?
The executive accountable for the program and the leaders of the functions affected, because employees want to hear from the people who decide about their work. A technology leader can explain the systems briefly, but a town hall run by the technology function reads as a project update rather than a commitment.
05What should happen after an AI town hall?
Written answers to everything asked, published where everyone can see them; answers to open items by the dates promised; manager conversations in affected teams; and a follow-up session that reports what actually happened. Town halls without follow-through make the next one worse than useless.
Continue exploring
Related capabilities
Start with the hard problem
Need the outcome owned, not merely analyzed?
Tell us where delivery is constrained. WeтАЩll map the fastest credible path from intent to verified production.