How Should a CEO Explain an AI Programme to Employees?

Employees need to know what an AI programme will do to their work, who controls it, and when leaders will decide what happens next. A broad promise about productivity leaves those questions open. A CEO should explain the first workflow in plain language, state what the company has decided about roles, and give staff a way to challenge the plan.

Settle the facts before the announcement

Before the all-hands meeting, ask the pilot owner to write down the task, affected team, start date, data the tool may use, actions it may take, and the person who can stop it. Agree on what result would justify wider use. Tell managers and the people who will work with the system first, so they can ask questions before a company-wide announcement.

Separate decisions from open questions. If the board has approved a pilot but has not decided whether it will change staffing, say exactly that. If leaders have a cost reduction target, don't hide it behind a promise that AI will only remove dull work. Staff will judge the message against the decisions that follow it.

Say exactly which work changes

Name the current task and the new handoff. In a customer support pilot, the tool might gather an order record and draft a reply from approved policy documents. An agent checks the draft before sending it. A supervisor still decides refund exceptions. This description tells a support agent where their judgement remains in the process and where the company expects the tool to help.

State which data staff may enter, where the output goes, and what the system cannot do. Give staff an approved route for reporting a wrong answer or a privacy concern. NIST's AI Risk Management Framework calls for clear human oversight and named responsibility for AI risks. A CEO's message should make those controls visible to the people using the system.

Answer the question about jobs

Employees will ask whether the programme changes roles, hiring, pay, or headcount. Answer each question with the decision you have today. Don't promise that no jobs will change unless the company has made and can keep that commitment. When a decision is still open, name who will make it, what evidence they will use, and when the affected team will hear from them.

The International Labour Organization's 2025 research finds that many jobs exposed to generative AI are likely to change in their tasks. That finding cannot tell employees what their own employer plans to do. Be direct about your company's goals and give people time and support to learn the changed work. If a role decision has already been made, tell the affected people through a direct, appropriate process before presenting it as a general update.

Give employees a real say

Ask the people who perform the task to show the exceptions that a tidy process map misses. Give them paid time to test real cases and report failures. Offer a named contact and a private route for sensitive concerns. Tell staff who reads feedback, when they will get a response, and how a reported problem can pause the pilot.

An OECD review of employer and worker surveys associates worker consultation and training with better reported outcomes. Consultation needs to change decisions when staff show that the system adds review work, misses context, or creates a new risk. Publish what the team changed in response, including cases where the pilot did not improve the process.

An example announcement

This short opening fits a support-team pilot. Change every claim about staffing, timing, and authority to match the decisions your company has actually made:

“We will test a tool that drafts answers to routine delivery questions using our approved policies. It will not send replies or approve refunds. Support agents will review every draft, and the support lead can stop the pilot. We will compare resolved cases, accuracy, and review time with the current process. We have not made a staffing decision from this pilot. The support team will see the first results before we decide whether to expand it. Please tell your manager or use the pilot feedback form when the draft is wrong, the policy is missing, or the new process adds work.”

Follow the opening with a written page that names the owner, the approved tools, the training time, the feedback route, and the date of the first results review. Managers should use that same page when staff ask what the announcement means for their own team.

Report what happened

Set the next update date before the pilot starts. Report the number of completed cases, correct outcomes, time spent reviewing drafts, errors, and staff feedback. Explain which problems the team fixed and which remain open. If the results don't support wider use, say so and stop or redesign the test.

Keep speaking with the affected team as work changes. A credible AI programme gives employees a clear task boundary, honest answers about roles, and evidence that leaders will change course when the process fails. The CEO's first message starts that agreement; the decisions after it earn trust.