There is nothing wrong with a leader saying AI will change the way that their organisation works.
It will.
But there is a big difference between setting out a thoughtful AI strategy and making a bold public declaration that makes your people sound like the problem.
That is the trap Duolingo fell into.
In April last year, its chief executive, Luis von Ahn, announced that the company was becoming “AI-first”.
The memo set out a series of firm expectations: gradually stopping the use of contractors for work AI could handle, looking for AI skills when hiring, assessing AI use in performance reviews, and only adding headcount if teams could show they could not automate more of their work.
That may have sounded decisive in the boardroom. Outside it, the message read very differently.
To employees, contractors, customers and commentators, it sounded less like a technology strategy and more like a warning. AI first. Humans second.
There was backlash, concern and, eventually, a walk-back.
Von Ahn later said the memo had not provided enough context. Duolingo also moved away from assessing employees on AI use in performance reviews, with the chief executive acknowledging that the real test should be whether people are doing their jobs well, not whether they are using a particular tool.
The “lack of context” defence is a familiar line when a bold statement starts creating awkward headlines. But there is a problem with blaming missing context.
The original words still exist.
And once they have shaped the story, the explanation rarely travels as far as the concern.
The problem is not that organisations are exploring AI.
The problem is that some leaders are over-claiming before the work is properly understood.
AI can improve many tasks. It can speed up content production, help with analysis and reporting, support customer service, summarise information and reduce repetitive day-to-day tasks.
But it is not a magic replacement for judgement, empathy, creativity, trust, experience or accountability.
When leaders talk as though it is, they make a promise the technology may not be able to keep.
At first, the bold message may sound impressive. It may create headlines. It may reassure investors that the organisation is serious about change.
But it also creates a promise that is hard to keep.
If the technology does not deliver as expected the organisation must explain the gap.
If jobs still need to be filled, it must explain the hiring.
If people have been cut and then quietly rehired under different titles, it must explain the reversal.
If quality drops, it must explain why speed was prioritised over care.
And if employees feel reduced to a cost line, it must rebuild trust.
That is the climbdown problem.
A cautious AI strategy can be updated.
An over-confident AI declaration must be retracted.
There is a big difference.
Duolingo is not alone.
Gartner has predicted that, by 2027, half of companies that cut customer service staff and attributed the reduction to AI will rehire people to perform similar functions, often under different job titles.
That is a striking forecast.
It suggests that some organisations may be discovering the hard way that replacing people with AI is much easier to announce than it is to make work in practice.
There is a communications lesson here.
If you tell the world AI has allowed you to reduce headcount, then later need people back, that is not just an operational correction. It is a credibility problem.
Were you wrong about AI?
Were you too quick to cut?
Did you understand the work your people were doing?
Did you put short-term optics ahead of long-term service?
Those are not comfortable questions for any leadership team.
They become much harder to answer if the original announcement was full of certainty.
The more dramatic the claim, the harder the retreat.
One of the recurring mistakes with AI announcements is treating internal and external communication as different worlds, when they are the same world with different audiences watching at the same time.
A memo to staff can become a LinkedIn post.
An all-hands meeting can become a screenshot.
A performance review policy can become a headline.
And a chief executive quote meant to project confidence can become the phrase employees repeat when trust starts to fade.
That does not mean leaders should say nothing.
It means they should write and speak as if every audience is listening simultaneously, because they probably are, and each one brings a different fear.
Employees worry about replaceability.
Customers worry about poorer service.
Investors worry about whether the savings are real.
Journalists look for the human impact, and regulators will examine fairness, accountability and risk.
The same sentence has to work for all of those audiences.
That is why "AI-first" is risky language.
It may be intended to signal ambition, but it can also suggest the organisation has moved people into second place. Once that perception forms, any later attempt to explain the context tends to sound like damage control.
This is not about hiding your AI strategy. It is about communicating it with more judgement.
First, avoid slogans that make people feel secondary.
“AI-first” may sound neat, but it invites the obvious question: where do humans come in?
Better language focuses on outcomes, customers, colleagues and the work itself.
Second, be clear about what AI will and will not do.
If AI is there to support people, say so clearly. If some roles are likely to change, be honest about that too.
People do not need vague reassurance that everything will be fine.
They need detail, candour and leaders who are visible enough to answer difficult questions.
Third, be careful not to confuse AI use with good performance.
The real question is not whether someone used AI. It is whether the work improved.
If people are rewarded for using tools rather than achieving better outcomes, they will use tools for the sake of it. That is not transformation. That is theatre.
Fourth, test the message with different audiences before it goes public.
How will this sound to someone worried about their job?
How will it sound to a customer who values human service?
How will it sound in a headline?
How will it sound if you have to reverse part of the strategy in six months?
Fifth, leave yourself room to learn.
AI is moving quickly. No leadership team has perfect answers. So do not communicate as if you do.
It is far stronger to say, “We are testing where AI can improve work, and we will keep people involved in those decisions,” than to imply the future has already been settled.
The organisations that communicate AI well will not be the ones with the loudest announcements.
They will be the ones that can explain the human impact with clarity and care.
They will show how AI supports better work, not just cheaper work.
They will talk about skills, roles, customers and responsibility.
They will make managers part of the conversation, not leave them trying to interpret a chief executive’s memo after the headlines have broken.
And they will avoid creating the kind of public promise they later have to row back from.
Because when leaders over-claim on AI, the risk is not just that the technology fails to live up to the message.
The bigger risk is that people stop believing the messenger.
That is the communications trap.
AI may well change your organisation.
But if your message makes people feel disposable, the problem is not a lack of context.
It is a lack of judgement.
Media First helps organisations navigate high-stakes communication challenges.
We combine strategic counsel with immersive training to build confident leaders, resilient teams and stronger reputations. To discuss how we can support you, contact iain@mediafirst.co.uk