Key takeaways
- Employees are already automating their own tasks with AI—the open question is who keeps the saved time.
- AI is reshaping work more than deleting it; most leaders plan to hire, not fire.
- Without a clear policy, people hide productivity gains to avoid being given more work for the same pay.
- Design a 'time dividend' that splits saved hours between the business, the employee, and shared training.
- Trust and transparency scale individual automation hacks; surveillance drives them underground.
There is a new kind of invisible labour happening inside almost every company right now. Employees are using AI to automate parts of their own travail—scripting repetitive tasks, drafting reports in seconds, cleaning data that used to eat an afternoon. The productivity gain is real. The awkward question nobody has answered is simple: who owns the time that was saved?
This matters because the usual narrative—AI will replace jobs—is turning out to be wrong, or at least incomplete. In the United States, close to nine CEOs out of ten say they plan to hire more people despite the rise of AI, not fewer. The technology is reshaping what work looks like far more than it is deleting it. And that reshaping is now happening bottom-up, by employees themselves, often without anyone in management noticing.
The grey zone of self-automation
When an employee builds a small automation that turns a two-hour task into a two-minute one, three things can happen with the recovered time. They can absorb more work. They can quietly keep the margin for themselves. Or they can surface the gain so the organisation can redeploy it. Most companies have no policy covering any of these outcomes, so the default is the second one—private, undeclared productivity.
We see this constantly with clients. The issue is not laziness; it is the absence of a social contract around automation. If employees suspect that every hour they save will simply be refilled with more tasks at the same pay, they have a rational incentive to hide their gains. That is how companies end up with enormous latent productivity they can neither see nor capitalise on.
Why transparency beats surveillance

A related tension is emerging in academia, where PhD students have started voluntarily disclosing their use of AI—not because rules require it, but to protect themselves against future accusations. That instinct is telling. In the absence of clear policy, people manage their own risk by over-declaring or under-declaring. The lesson for businesses is that clarity is a feature, not a constraint. If you define what good AI use looks like, people stop hiding and start sharing.
The alternative—monitoring keystrokes and measuring output more aggressively—backfires. It pushes the most capable automators to conceal exactly the innovations you most want to scale. Trust, not surveillance, is what turns individual hacks into organisational capability.
A fair framework for the time dividend

At Evocative, we help organisations design what we call a time dividend: an explicit agreement on how automation gains are split. There is no single correct ratio, but the principle should be a genuine three-way share.
- The business reclaims part of the time to grow capacity or margin—legitimately, since it funds the tools and the risk.
- The employee keeps part of the gain as autonomy, deeper work, or reduced overtime pressure.
- The collective gets part reinvested into training, so AI literacy spreads instead of concentrating in a few quiet experts.
This reframes automation from a threat into a negotiated upgrade. It also connects to a broader cultural fatigue: the pressure to constantly optimise ourselves, to treat every hour as a project to be improved, has turned personal development into a second unpaid job. People are tired of being their own relentless manager. A company that offers a clear deal on automation gains—rather than silently expecting more from everyone—stands out in a labour market where that exhaustion is widespread.
The policy you can write this quarter
You do not need a committee. You need a short, honest internal document that answers four questions: What AI tools are approved? What must employees disclose when they automate a task? How are saved hours accounted for? And what does the employee get in return? Publish it, test it on one team, and iterate.
The companies that win the next phase of digitalisation will not be the ones with the most AI licences. They will be the ones that resolved the ownership question early—who decided, out loud, that the time saved belongs to everyone, and built the trust to make that true. The automation is already happening in your organisation. The only choice left is whether you benefit from it in the open, or lose it in the dark.
Sources
- L'Insee estime que l'emploi salarié privé a perdu près de 26 000 postes — Journal du Net
- Les jours fériés tombent très mal en 2027 sauf pour les salariés dont le contrat est en forfait jour — Journal du Net
- Le développement personnel est devenu un travail de plus. Quand sommes-nous devenus nos propres patrons ? — Journal du Net
- Aux Etats-Unis, près de 9 PDG sur 10 prévoient d'embaucher davantage malgré l'essor de l'IA, voici pourquoi — Journal du Net
- IA : faute de règles, les doctorants l'utilisent à leurs risques et périls — Journal du Net
- Cette entreprise distribuera 200 milliards de dollars à ses salariés à condition qu'ils restent au travail — Journal du Net