Geyto Research · Juan Ferreiro · Essay · · 3 min read

AI training. What a course does not build

Teaching people to use an AI tool is necessary, but it does not change the work on its own. What the data says and what else adopting AI takes.

In brief: when an organization decides to "adopt AI", it is natural to start with a course. Teaching people to use the tool is necessary. But AI asks for something that training alone can hardly build: judgement about when to rely on what it produces, agreements about who reviews and who is accountable, and practice with real cases from the work itself. AI training matters; it is not enough.

Training arrives late and thin

The data shows that in many organizations AI training has not even arrived. In Grant Thornton's survey of 950 business leaders (2026), 34% of finance leaders name training as the most underfunded area of AI investment. Only 6% of executives rank change leadership and workforce enablement as an essential capability.

In Argentina, the CEPE Di Tella and Fundar survey found that only 13.7% of those who use AI at work received any training from their organization. 85.6% learned on their own.

Using the tool is not knowing when to use it

Even when training exists, teaching people to use the tool does not solve the hardest part. In the experiment by Dell'Acqua, Mollick and colleagues with 758 consultants, AI improved results on tasks that fell within what it does well and worsened them on a task that fell outside, even though it looked similarly difficult. The boundary was not obvious to the people using it.

That defines the kind of capacity AI demands. Knowing what to type is not enough. People need to recognise when a result that looks right is not, and that is learned with cases from their own work, not with generic examples.

Transfer depends on the job

Research on transfer of training, since the review by Baldwin and Ford, shows that whether what is learned gets applied at work depends on who learns, how it is taught and, with considerable weight, the environment where it will be used. With AI, that environment includes new questions: which tasks can be done with the tool, what gets reviewed, who is accountable for the result and which data can be used.

If those questions have no answer within the team, training remains individual knowledge. Each person decides on their own, and the work as a team does not change.

Where adoption is decided

In Grant Thornton's survey, executives say the people who need the most support are frontline employees (37%) and middle managers (30%). That is, the people who have to change how the work is done every day and those who coordinate that change. The same survey shows a gap in perception: technology leaders are five times more likely than operations leaders to say the workforce is fully ready to adopt AI.

That gap is a finding in itself. From the technology side, the tool is available. From the operations side, the work has not changed yet.

What to do about it

  • Train with cases from the work itself, including cases where the tool gets it wrong.
  • Agree within each team what can be done with AI, what gets reviewed and who is accountable.
  • Give middle managers an explicit role: they are the ones who sustain the change day to day.
  • Support people on the job after training, not only before launch.

More in Training matters, but it is not enough, in Why training does not always reach the job and in Geyto Learning.

Geyto Research publishes essays, evidence reviews and market notes on AI and technology adoption in organizations.

Frequently asked questions

Is it enough to train teams in AI?

It is necessary, but not sufficient. Training teaches people to use the tool; adoption also requires judgement about when to rely on it, team agreements about what gets reviewed, and time to practise with real cases on the job.

Are companies investing in AI training?

Less than they invest in the tool. In Grant Thornton's 2026 survey, 34% of finance leaders name training as the most underfunded area of AI investment. In Argentina, only 13.7% of those who use AI at work received training from their organization.

Who needs the most support to adopt AI?

In Grant Thornton's survey, executives point to frontline employees (37%) and middle managers (30%): the people who have to change how the work is done every day.

Sources

  1. Grant Thornton (2026). 2026 AI Impact Survey Report. Survey of 950 business leaders, February–March 2026.
  2. Avenburg, A., Bendersky, A., Judzik, D. et al. (2025). Encuesta Nacional sobre Adopción de IA en Argentina. Relevamiento Individuos, August–September 2025. CEPE, Universidad Torcuato Di Tella and Fundar.
  3. Dell'Acqua, F., McFowland III, E., Mollick, E. R. et al. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality. Harvard Business School Working Paper 24-013.
  4. Baldwin, T. T. and Ford, J. K. (1988). Transfer of training: A review and directions for future research. Personnel Psychology, 41(1), 63–105.

Share