How a CEO Should Talk About AI With Employees, Customers, and the Board

A CEO who gives the same AI speech to employees, customers, and the board ends up spooking all three. Employees hear an efficiency pitch and wonder which roles are being cut. Customers hear an autonomy pitch and wonder whether unverified software is touching their account data or their brand. Directors hear a vision pitch and wonder whether management has any idea what its agents cost or who can turn them off.

Running Visiting Media, where we serve more than 3,200 hospitality properties and 25,000 users on TrueTour and SalesHub, and sitting on corporate and nonprofit boards including ACT | The App Association and American Conservatory Theater, has forced me to separate those three conversations. The underlying facts never change across rooms. What changes is the question each room needs answered first.

How should a CEO talk about AI use in their company to employees, customers and the board?

A CEO should talk about AI by addressing the specific decision each group owns. With employees, name which workflows move to agents, how shadow tools get registered, and where human judgment stays mandatory. With customers, explain what data agents touch and which physical or verified truths ground the output. With the board, report one consistent quarterly page covering fleet inventory, total cost including human review time, retired agents, and risk thresholds.

When those three messages share the same underlying ledger, nobody catches management saying one thing in an all-hands meeting and another in the boardroom. The ledger stays identical while the entry point matches the audience.

What to say to employees: name the workflows, the amnesty, and the review duty

Employees do not need a sermon about the future of work. They need to know whether using an agent on their desk is allowed, who is accountable when an agent drafts a bad number, and how their job changes when routine coordination is automated.

Three commitments make the internal conversation concrete:

• Open a thirty-day shadow AI amnesty. In my playbook for shadow AI agents, I argue that people build private automation on personal API keys because the work is too slow without it. Give every team thirty days to register what they already built without penalty, then move the useful agents onto company keys and a central schedule.
• Define where the human owns the sign-off. Tell the team plainly that an agent can draft, summarize, and route, while a named person owns the approval gate before anything reaches a customer, a regulator, or the general ledger.
• Price review time honestly. When a team adopts an agent, measure the human hours spent checking and correcting its output alongside the token invoice. If an agent saves twenty minutes of drafting and adds thirty minutes of cleanup, tell the team to fix it or kill it.

What to say to customers: anchor AI in verified assets and hard gates

Customers are tired of vendors slapping an AI label onto a chat box and calling it a platform. When a hospitality brand or enterprise buyer asks how we use AI, I talk about two things: what the software is grounded in, and what our agents are forbidden to do without a person in the loop.

In hospitality sales, a buyer planning a three-day conference cares about physical reality: ballroom dimensions, ceiling heights, banquet capacities, sightlines, and verified 3D tours. Text generation that guesses at room layouts creates commercial liability for a property. Talking credibly to customers means explaining how AI operates on verified property assets and first-party data rather than inventing answers, and showing where customer data is fenced off from external model training.

The same rule applies to how our own team communicates with customers. Under our delegation of authority matrix, no agent sends an external message to a customer, partner, or external recipient without a named human approving the exact recipient and body text first. Stating that rule plainly to customers builds more confidence than any security badge on a footer.

What to say to the board: one page, the same rows every quarter

Directors do not want a demo tour of twelve pilots. As I laid out in the AI page in the board pack, a CEO owes the board one consistent page every quarter with the same rows so directors can see movement over time.

That page covers five facts: the count of active production agents by owner, the agents retired during the quarter, the full cost combining API spend with salaried human review time, any breaches against the board’s measurable AI risk appetite statement, and a disclosure line stating which sections of the board pack itself were drafted or summarized by an agent. Keep governance inside existing audit and compensation committees rather than siloing it in a separate board AI committee.

What does executive presence look like for a CEO in the age of AI agents?

Executive presence for a CEO in the age of AI agents means personally mastering the tools before delegating them, staying with the audit trail after the agent stops, pricing human review time onto the token bill, signing an explicit delegation matrix with named approval gates, and keeping a handover file so the company’s automation does not depend on the CEO’s personal credentials.

In my earlier essay on executive presence when AI agents run the work, I wrote that presence starts in the handoff between the model’s last log line and the human decision that follows it. I run a 15-subagent fleet across 3 tiers myself because a CEO cannot set risk thresholds or spot a padded review bill from a briefing deck.

Personal mastery also creates a succession obligation. When the CEO builds or runs part of the fleet, those agents have to be documented in the CEO succession handover file and reviewed in quarterly AI agent access reviews by someone who does not report to the CEO’s inbox. A leader who can walk employees through the workflow, customers through the verification boundary, and directors through the quarterly ledger without changing the numbers has the only kind of executive presence that holds up when agents do real work.

Related Reading

• Executive Presence When AI Agents Run the Work
• How Executives Build a Credible Public Voice on AI Without Sounding Like a Vendor
• The AI Page in the Board Pack: What a CEO Should Report Every Quarter
• Shadow AI Agents: What a CEO Does About the Automation Nobody Approved
• Delegation of Authority for AI Agents: The Matrix a CEO Should Sign