Leadership

When AI starts speaking for the leader

AI can improve a leader’s writing. It can also make it harder for people to tell whose judgement they are hearing.
By Wadda Salah Eldin · · 7 min read

Essay

A leader writes a difficult message about the company’s AI strategy. She knows people are anxious. She drafts it, then uses a generative AI tool to make the language clearer.

The result is polished. It may even be better written. But the team is not only reading for information. They are trying to understand what their leader believes, what she has decided and whether she understands what the change will ask of them.

If the tool has supplied the certainty, empathy or reasoning, the message may no longer answer those questions.

The issue is not whether AI touched the words. It is whether the leader still owns the judgement inside them.

People listen for the person behind the message

Leadership communication is not simply the transfer of information. People compare a leader’s words with the decisions that follow, the trade-offs the leader accepts and the behaviour they see over time.

AI can help someone organise a thought or remove unnecessary complexity. It can also produce language that sounds more certain, more considered or more empathetic than the person using it. That creates a gap between the voice in the message and the person who must stand behind it.

The research cited below points to a narrow but useful concern. In the tested context, high AI involvement reduced perceived empathy and credibility. It does not prove that every AI-assisted message will be distrusted. It does suggest that how a leader uses the tool matters.

Three ways a leader can use AI

1. To examine their own thinking. A leader can ask AI to challenge an argument, identify missing assumptions or show where a message may be unclear. The tool helps the leader think, but does not decide what the leader believes.

2. To perform a voice they have not earned. A leader can ask AI to supply empathy, confidence or conviction that is absent from the underlying decision. The words may improve while the leadership becomes less honest.

3. To improve work they still own. A leader can draft the substance, use AI to test or refine it and remain responsible for every claim, decision and consequence. Where the use is material, the leader can explain it without embarrassment or evasion.

These are not rigid categories. The same person can move between them in a single afternoon. The useful distinction is whether AI is helping the leader see and express their own judgement, or quietly replacing it.

A better test than “Did AI write this?”

Ask four questions before a consequential message goes out:

  • Can I explain the decision in my own words without the draft in front of me?

  • Does the message acknowledge what is still uncertain?

  • Have I checked every factual claim and accepted responsibility for it?

  • If someone asked how AI was used, could I answer plainly?

Disclosure is not a ritual that belongs on every sentence. It matters when AI has materially shaped the reasoning, recommendation or communication, or when people need that context to judge what they are being told.

Keep the difficult parts human

Leaders should be especially careful when the message carries bad news, asks people to trust an uncertain plan or explains a decision that affects someone’s role. Those are precisely the moments when polished language can hide the absence of a settled position.

AI can help with structure and clarity. It cannot take responsibility for the decision. It cannot sit in the room when people challenge the trade-offs. It cannot repair a promise the organisation fails to keep.

Use the tool to improve the message. Do not use it to manufacture the leader behind it.

Evidence

Sources and limits.

Last materially reviewed
September 2026

Basis
Interpretive essay using experimental research on AI involvement in leadership communication.

Status
The cited study supports a narrower claim: high AI involvement can reduce perceived empathy and credibility in the tested context. It does not establish a universal response.

Source
Sedefoglu-Ulucak, Ohly and Schmelz, “Empathy in leadership communication,” Computers in Human Behavior, 2026 (opens in a new tab)

Apply the idea

Where must a leader’s own judgement and language remain visible?

Discuss a live situation with us