Managing Teams with Artificial Intelligence: the Risks Nobody Names

An operations manager recently described an uncomfortable situation. His team started receiving recommendations from an AI system about daily priorities, and within three weeks people had stopped discussing priorities in meetings altogether. They simply followed what the screen said. Nobody had decided that consciously. It just happened.

This kind of situation repeats across many teams that adopt AI tools without preparing the process around them. The technology arrives to solve a visible problem, saving time, cutting calculation error, and without anyone noticing, it starts replacing a practice the team did not even realise it valued: discussing before deciding.

What risks come with managing teams using AI tools?

The most discussed risk is technical: model errors, outdated data, recommendations that miss the context. That risk is real, but it is not the most dangerous one. The risk that rarely gets named is relational. A team that starts treating a system's recommendation as the final decision stops exercising the muscle of discussing, disagreeing and adjusting together. That muscle, once lost, is expensive to rebuild.

There is also a risk of diffused responsibility. When something goes wrong after following an AI recommendation, it is common to hear "that's what the system said." In a mature team, that sentence does not exist, because the responsibility for validating a recommendation still belongs to whoever applies it, not to the tool that suggested it.

How do you keep the team's trust when AI decides for them?

Trust holds when the team experiences AI as a support instrument, not as an authority that replaces conversation. That is not secured with a motivational speech. It is secured with practice: a manager who keeps asking "what do you make of this recommendation" before applying it, even when it looks obvious, keeps alive the discussion that the tool could otherwise extinguish out of convenience.

When a team senses that a decision about their work, a reorganisation, a change of priorities, was made by a dashboard with nobody explaining why, trust breaks fast and repairs slowly. Human explanation remains irreplaceable, even when the data is solid.

Does AI increase or reduce the manager's workload?

Both, at different moments. It reduces the operational load of gathering and cross-referencing information, work that used to take hours of preparation before a team meeting. It increases the relational load, because the manager now also has to manage the team's reaction to recommendations that did not exist before, and decide, case by case, when to follow that recommendation in front of the group and when to overrule it.

Managers who only count the first half of the equation tend to underestimate the toll the second half takes. The workload does not disappear. It changes shape, and the new shape demands skills that many managers never had to train before.

How do you know if your team has the maturity to absorb more AI?

Evomatrix, the decision maturity diagnostic used in Hélder Teixeira's work, helps answer this question before a company deepens its reliance on automated systems. It measures how the team already decides under pressure, where it discusses well and where it avoids conflict, and uses that picture to predict which kind of error AI will amplify first if it is introduced without preparation.

A team that already discusses disagreements openly will keep doing so with an AI tool alongside it, treating the tool as one more source of information. A team that avoids conflict will use the tool's recommendation as an excuse to avoid that conversation for good, which looks like efficiency and is, in practice, a worsening symptom.

This diagnostic does not replace the decision to adopt an AI tool or not. It calibrates the pace of that adoption and the kind of support the team will need during the transition, rather than treating implementation as a purely technical project run by the IT department.

What signs show a team is losing depth because of technology?

One clear sign is a drop in the number of questions asked in meetings after data gets presented. Another is less time spent discussing why a decision was made, replaced by an almost exclusive focus on what to do next. A third sign, harder to spot from outside, is a decline in individual initiative, when people start waiting for the next recommendation instead of bringing their own observations to the table.

None of these signs shows up overnight. They set in slowly, which is exactly why an attentive manager needs to look for them actively, rather than waiting for them to become obvious enough to show up in a climate report.

How do you correct course after spotting these signs?

The first step is naming the pattern out loud in front of the team, without blaming anyone. Saying, with concrete data, that meetings have had less discussion since the tool came into use already changes the dynamic, because it makes visible something that had been happening unnoticed.

The second step is deliberately redesigning the decision process to include a moment of discussion before any AI recommendation gets applied, even if that moment takes only a few minutes. It is not about distrusting the tool. It is about keeping alive the practice that makes a team think together, a practice no technology should replace unless someone consciously decides that.