In the short term, AI changes work more than it changes headcount by itself. Some activities go from hours to minutes; people do not disappear automatically, but they do different things and take on new responsibilities.
What the evidence says about AI and work
It helps to separate three levels that often get mixed up:
- task exposure: how much an activity can be assisted or automated;
- potential automation: what is technically feasible;
- actual job loss: what really happens in the organization, which depends on managerial choices, market conditions, and reorganization.
According to the ILO on GenAI’s impact on work, exposure is mainly about transforming tasks rather than an immediate total replacement of roles. The European framework from OpenAI on the employment transition proposes change archetypes (growth, automation, reorganization, and less immediate impacts): a useful map, not a certainty about jobs created or eliminated.
In short: many activities transform; few binary narratives hold up against the data.
New operational responsibilities emerge
When AI enters processes, someone must:
- check the quality of outputs;
- design flows, prompts, and controls;
- manage company knowledge so it is usable;
- validate decisions and keep human accountability.
These are not trendy titles: they are concrete tasks. If nobody owns them, AI produces speed without reliability.
Why the managerial message changes adoption
If the message is “we are cutting costs,” you get resistance. People go on the defensive, see AI as a threat, and may even obstruct its use, sometimes unconsciously.
If the message is “we are increasing capacity and quality,” you get more collaboration. AI is read as a tool to do better work, not as a shortcut to remove people.
Honesty matters: over the long term, some roles will change significantly. That has happened with other transformative technologies; historical analogy is context, not proof. The useful response is not principled resistance, but adapting and training people for the new tasks that emerge.
What to do in practice in SMEs
- Communicate the purpose clearly: more quality and capacity, not only cuts.
- Involve the people who live the process in designing the solution.
- Invest in concrete training on the use case, not only generic webinars.
- Assign ownership for quality, knowledge, and supervision.
- Measure time, quality, errors, and trust, not only theoretical savings.
Without this work, even a good model remains a technical project with weak adoption.
FAQ
Which jobs are most exposed?
In general, repetitive, highly standardized activities based on clear rules. Work that requires judgment, negotiation, relationships, and accountability typically stays human, even when support tasks change.
How should employee fear be managed?
With transparency, involvement, and concrete results on a real process: a useful pilot builds more trust than a theoretical presentation.
Do SMEs have the resources to train?
Yes, if they train on the specific use case and use scoped pilot projects as the learning context.
Will AI eliminate net jobs?
There is no single answer. In the short to medium term, the nature of work often changes more than the overall number of people; the outcome depends on organizational choices and reskilling.
Sources
- ILO: Generative artificial intelligence and work: global index: task exposure and the prevalence of transformation.
- ILO: Impact of GenAI on employment and organization: evidence and limits on labor impact.
- OpenAI: Mapping AI opportunities for the European workforce: transition archetypes, not certain forecasts.
Explore the series
- Data security and Shadow AI: why bans are not enough
- Where to start with AI in the company
- Document management and AI
If you want to introduce AI on a real process while accompanying people, responsibilities, and training—not only the technology—we can design an adoption and governance path together. Write to us at info@zendata.it or visit zendata.it.
Pietro Ciattaglia, CEO of Zendata AI, Rome