Can Middle Managers Be Trained Like Executives?

1. Two levels, two approaches to training.

At the top of the organizational chart, professional development is handled on a case-by-case basis, in an ultra-personalized way. A coach, ten sessions. An executive education program where you work on your own cases. An assessment center whose results are discussed one-on-one. But obviously, it is expensive. As a benchmark, INSEAD's Advanced Management Programme costs €47,500 for four weeks, excluding accommodation.

In the middle of the organizational chart, professional development (because of the volume of learners) is standardized, limited, industrialized, so to speak. Fifteen people in a room, a trainer, slides, a role-play between two coffee breaks. Obviously, the benefits are not the same (see our article on the 10% myth).

What makes leadership training so effective is its level of personalization, its duration, and its ability to provide hands-on practice.

2. How much would executive-level training cost if it were applied to middle management?

Let's take a company with 5,000 employees, with one manager for every eight employees: around 600 middle and frontline managers (what is commonly referred to as middle management). If we gave them the same development path as executives, we can add up the costs using conservative lower-end estimates:

• Coaching: 10 sessions at €250 to €500 per session. As a benchmark, the ICF reports an average of $277 per session in Western Europe, across all specialties; we use a higher figure here for executive coaching.

→ €2.5k to €5k

• Short executive education program.

→ €8k to €20k

• Assessment or 360-degree feedback debriefed one-on-one.

→ €1k to €3k

€11.5k to €28k per manager. Let's say €15k for the sake of a simple round number. For 600 managers, that's €9M.

Now let's look at the budget currently available. Training Magazine's 2025 annual survey of U.S. companies with more than 100 employees reports $874 in training expenditure per learner, of which 13% is allocated to managers and supervisors.
So, assuming a rough approximation, an annual budget for a 5,000-employee company would be around €500k.

The budget available for middle management is therefore 18x too small to offer them high-end development programs.

3. The irony: those who get the most out of training are the least well served.

We could take comfort in thinking that the effort is at least concentrated where it generates the most value. But the data suggests rather the opposite.

The meta-analysis by Lacerenza, Reyes, Marlow, Joseph & Salas (2017) compiles 335 studies and 26,573 participants. It distinguishes three levels of leaders and measures the effect of training on behavioral transfer, i.e. changes in behavior at work:

• Frontline managers (in direct contact with teams): strong transfer (58 studies, 3,140 people).

• Middle managers: moderate transfer (44 studies, 2,413 people).

• Senior leaders: very low transfer (19 studies, 995 people).

The further down the organizational chart you go, the greater the impact of training on behavior.

Two caveats. First, these are subgroups within a meta-analysis, not a trial comparing the different levels under otherwise identical conditions: studies on frontline managers may not measure transfer in exactly the same way. Second, the meta-analysis does not explain why. A reasonable hypothesis is that a new manager starts further behind, and every skill they acquire is immediately visible in their day-to-day work.

One stubborn fact remains. The population that responds best to training is the one receiving the poorest format.

4. What AI changes.

Let's go back to the executive development path. What makes it expensive? Not the content: the management principles taught to executives and frontline managers are largely the same. What is expensive is the time of a qualified human working with a single person. Coach, actor, assessor, facilitator.

This is exactly what AI can reduce. Roughly speaking:

• One hour with a qualified professional.

→ €250 to €500

• One hour of simulation powered by a language model, in inference costs.

→ €1 to €5

And the trend is moving in only one direction: for equivalent performance, a16z measures a tenfold decrease in inference costs every year since 2021 (while noting that there is no guarantee this pace will continue).

Caution: this calculation concerns the marginal cost of one hour of practice, not the cost of a platform. Designing credible scenarios, calibrating them, and measuring what happens all have a cost. But this is a fixed cost that is divided by the number of managers rather than multiplied by it.

In 2025, Kestin et al. published a randomized trial at Harvard (194 physics students): students supported by an AI tutor learned more than in an active-learning class, in less time (49 minutes at the median versus 60). The authors themselves point out that the result concerns an initial exposure to concepts and that they make no assumptions about skills requiring complex and integrative thinking.

AI tutors obviously have their limitations; AI does not replace a coach. However, it makes individualized practice more affordable.

And what middle management is primarily lacking is practical training. A place to practice, where every training session is highly personalized.

This is where AI can have a real impact. We still cannot provide every middle manager in a company with coaching sessions and the level of support offered by the best business schools. But we can provide them with personalized tools that replicate the principles of leadership training, at a cost and with logistics compatible with the number of people who need to be trained.

5. What does this change for an HR department?

The practical consequence is not “let's give everyone an executive development program.” It is to extract the active ingredients of executive development and make them available further down the organizational chart.

Three ingredients are particularly important, all documented by Lacerenza et al. (2017):

• Practice: working on realistic situations that people actually encounter.

• Feedback: immediate, precise feedback on what the person actually did.

• Iteration: multiple sessions spread over time, to repeat and reinforce learning.

Until now, these three ingredients required an hour of human time costing several hundred euros. So organizations had to choose who would get access to them. And logically, they chose those whose every decision carries the greatest weight.

That choice is no longer imposed by the budget. Everyone can have access to it.

Conclusion

We have never trained middle management like executives, not because we thought it was unnecessary, but because it was 20x above companies' budgets.

This constraint held as long as individualized practice required one human per learner. It gives way as the cost of an hour of simulation continues to collapse.

At YouMight, we are building a platform to de-risk management transitions: a simulated organization, with its teams, channels, and memory, where every new manager can practice and receive feedback as often as an executive. Not a privilege reserved for the executive committee. Training for everyone who manages people.

Sources

• Lacerenza, Reyes, Marlow, Joseph & Salas (2017), Journal of Applied Psychology: meta-analysis of 335 studies, transfer by hierarchical level, effects of feedback and spacing
• Bloom (1984), Educational Researcher: “The 2 Sigma Problem,” finding group methods as effective as individual tutoring
• VanLehn (2011), Educational Psychologist: effect of human tutoring at d = 0.79, intelligent tutoring systems at d = 0.76
• Jones, Woods & Guillaume (2016), Journal of Occupational and Organizational Psychology: meta-analysis of 17 studies on workplace coaching
• Kestin, Miller et al. (2025), Scientific Reports: randomized trial with 194 students, AI tutor versus active learning class
• International Coaching Federation (2023), Global Coaching Study: 14,591 responses, average session fee by region
• Training Magazine (2025), Training Industry Report: expenditure per learner and budget allocation by population
• INSEAD, Advanced Management Programme: 2026 fees
• Appenzeller (2024), a16z: “Welcome to LLMflation,” decline in inference costs at equivalent performance