Everyone currently managing AI anxiety in their workforce is rerunning an experiment whose results I already have. During the paper-to-computer transformation at a major national retailer, the people whose work was being digitized carried a quiet fear. That they’d lose their jobs. Or that they wouldn’t be able to do their jobs in the new system. It was a back-of-the-mind fear, a murmur. And it drove real resistance anyway, the subject of the companion article on the paper wars.
Today the fear has a new object and a much higher volume. People are freaked out, and in many cases rightly so. This technology reaches further up the skill ladder than the computers ever did. But the fear is the same species, and I’ve watched what kills it.
Why did retraining work?
We retrained the paper workforce into the new system, and that solved the problem. It didn’t manage the problem. It solved it. The mechanism is worth taking apart, because it wasn’t the skills alone.
Retraining answered the actual question. The fear was “do I exist on the other side of this.” A training program answers that question with money and hours instead of words. Companies don’t invest in training people they intend to discard, and employees know it. Every hour of instruction was evidence about their future that no town hall could provide.
And retraining converted the threat into a skill. The person who feared the system became the person who could run it. That removes the fear and inverts the incentive. The retrained clerk now had a stake in the new world working.
How does the retraining lesson translate to AI?
From the podcast
Trevor Sumner reached for the same historical pattern on Leaders and Their Stories, naming two technologies that were supposed to end professions: “The typewriter and the personal computer. We talked about how that was going to bring all the secretaries and note takers and typists out of work. And it created millions of jobs.” The jobs that came back weren’t the same jobs, and retraining is what got people into the new ones.
Companies don’t invest in training people they intend to discard, and employees know it.
Telling everyone AI won’t take jobs is the wrong translation, and it’s a promise people see through inside a week. Some jobs will change beyond recognition and some will go.
A leader who pretends otherwise burns the credibility the retraining depends on, and I have no patience for the executives doing it right now. The version that works has three parts, and we ran all three in the paper era without knowing we were following a method.
First, be specific about what changes. Vague reassurance amplifies fear; specificity shrinks it to a size people can plan around. Second, invest visibly in retraining toward the new shape of the work. AI-augmented roles need operators, reviewers, and people whose judgment directs the tools. Your current workforce holds the domain knowledge those roles require. Third, let the retrained become the proof. In our transformation, the turning point came when the first wave of paper veterans worked the new system competently, in full view of everyone still afraid.
| Move | What to do | Why it works |
|---|---|---|
| Skip false promises | Don’t tell everyone AI won’t take jobs | People see through it inside a week |
| Be specific | Say exactly what changes | Specificity shrinks fear to a size people can plan around |
| Invest visibly | Fund retraining toward the new shape of the work | Companies don’t train people they intend to discard |
| Retrain toward the domain | Move domain experts into operator and reviewer seats | Their judgment catches the machine confidently getting it wrong |
| Let the retrained prove it | First wave works the new system in full view | That was the turning point in the paper era |
Reassurance is free and everyone knows it. Retraining is a commitment with a budget attached, and that’s why it works.Share on X
What should you retrain toward for AI?
“Invest in retraining” fails as advice unless someone names the destination, so I’ll name the shape I saw then and the shape I see now. In the paper era we didn’t retrain clerks to be programmers. We retrained them to run the new system in their own domain. Their domain knowledge was the asset that made retraining worthwhile. The person who understood twenty years of payroll exceptions became the person who could tell when the payroll system was wrong, a skill no new hire and no vendor could supply.
The AI translation is direct. Skip the fantasy where everyone learns to build models. Move your domain experts into the seats the new tools create around them: the operator who directs the tool, the reviewer who judges its output, the person who knows twenty years of exceptions and can catch the machine confidently producing the twenty-first. AI raises the value of exactly the judgment your veterans hold, as long as you retrain them into positions where that judgment gets applied to the machine’s work.
Fail to do that, and you’ll discard the judgment with the job, then rediscover its price one confident machine error at a time.
My take
Retrain your veterans into the seats AI creates around them. Promise nobody their job is safe, because people see through that inside a week. Be specific about what changes. Move your domain experts into operator and reviewer roles, where twenty years of exceptions lets them catch the machine confidently getting it wrong. Let the first retrained group prove it works in front of everyone still afraid. Throw that judgment away with the job and you’ll pay for it one confident error at a time.
What happens if you get AI retraining wrong?
An organization that deploys AI over a fearful workforce gets what we would have gotten without the retraining: quiet sabotage of adoption, parallel processes kept alive in secret, and its most experienced people spending their energy on self-preservation instead of the mission. The technology budget buys nothing without the adoption, and the adoption is purchased with credible answers to the survival question.
The injustice is who pays for that mistake. The executive who bought the tool keeps the job and writes the lessons-learned memo. The clerk with twenty years of payroll exceptions in her head gets the severance letter, and the company throws away the one person who could have told it the machine was wrong. That’s a leadership failure, and I’m watching companies make it right now.
I write about this from both sides now: as someone who led the earlier transformation, and as an author and ghostwriter working daily with AI in a craft everyone said it would replace. My conclusion from both vantage points is the same, and I’ve written it before: augmented beats replaced, every time, and retraining is how a company chooses augmentation on behalf of its people.
For more from this series, see The Digital Transformation Hub: real transformations, lived from the inside, decades before the term existed.
Workers fear AI the way they feared computers. The Birth of the Augmented Human is my book on what goes right when people use AI properly, as a force multiplier for thinking.
Frequently Asked Questions
- Trevor Sumner on AI, Demographics, and the Disruption of Everything, Audible (Leaders and Their Stories)
