Reshare — content from Benoit Raphael
A study of 12,000 employees shows it: with AI, we talk more to our colleagues. Good news or a trap?
Researchers at Bocconi University in Milan tracked the meetings, calls and emails of 12,040 employees at a multinational company during the rollout of Copilot, Microsoft’s AI assistant. Three months after adopting the tool, users spend about twenty more minutes per week exchanging with their colleagues. Especially in meetings and on the phone.
But the real issue is not so much the time spent as the people we talk to. And above all, what about!
Copilot users work with nearly three more colleagues than before, people who know each other little.
Another survey, conducted by two researchers at the University of Berkeley (2026), observed it for eight months in an American technology company. Designers start coding. Researchers take on engineering tasks.
Imagine what that changes in a discussion: you can arrive with a first mockup, a piece of code, a proposal that you would not have been able to produce alone. Your colleague has something to examine, to discuss, to improve with you.
But this same survey also shows a real trap:
Engineers spend more time correcting the work that their colleagues did with AI. The machine produces faster than we understand.
If the twenty minutes of the Bocconi study are used to reread, verify or correct… the work has simply shifted to the other collaborators.
This gives a fairly concrete question to ask when we save time with AI: how much work remains to be done for the person to whom we pass the result?
Tell me what you think in the comments.
LinkedIn Sources
- Profile: Pierre Jean Duvivier dit Sage (LinkedIn)
- Shared author: Benoit Raphael (LinkedIn)
The post A study of 12,000 employees shows it: with AI, we talk more to our colleagues. Good news or a trap? appeared first on les plumes qui mordent.
Frequently Asked Questions
FAQs A Study of 12000 Employees Shows It With AI People Talk to Their Colleagues More Good News or a Trap
1 What does this study actually say
It found that when employees use AI tools at work they tend to communicate with their colleagues morenot less The study surveyed about 12000 employees
2 Who conducted the study and where can I find it
The headline comes from a specific study of 12000 employees but you should check the original source for the authors institution and publication details before citing it Headlines often simplify complex findings
3 What counts as AI in this study
Usually tools like chatbots writing assistants coding helpers and AI features built into workplace software The exact definition depends on how the study measured AI use
4 Why would AI make people talk to colleagues more
A common explanation AI handles routine tasks faster freeing time for collaborationand people often turn to colleagues to check refine or apply what AI produces
5 Is this good news
It can be More communication can mean better teamwork knowledge sharing and faster problemsolving But more talking isnt automatically betterit depends on what the conversations are about
6 So wheres the trap
Possible downsides include more interruptions more time spent verifying AI output with others AIinduced busywork and reliance on colleagues to fix AI mistakes More communication can also mean more meetings and less focused work
7 Does this mean AI reduces isolation at work
Possiblyif the extra communication is genuine collaboration But messaging more often isnt the same as feeling more connected The quality of interaction matters
8 Could the finding be correlation rather than causation
Yes Maybe people who already collaborate a lot are more likely to adopt AI rather than AI causing more collaboration Good studies try to control for this but its worth checking
9 How was talking more measured
Likely through selfreports communication logs or meetingmessage data Each method has limitsselfreports can be biased and logs dont capture conversation quality
10 Does the effect apply to all jobs
Probably not equally Knowledge workers teams using collaborative software and roles with lots of writing or analysis may
