The hidden cost of AI efficiency

A study of 187,000 developers on GitHub, before and after the introduction of GitHub Copilot, shows more than just productivity gains. It shows the transformation of the very nature of work. As MIT’s Frank Nagle puts it, “Generative AI has given people the ability to do more of what they want to do and less of what they have to.” It sounds like a dream come true for efficiency. The numbers from the study are impressive. The time spent coding has increased by more than twelve percent. Project management activities fell by almost twenty-five percent. But there’s another number, one that should give us food for thought: collaboration between developers has decreased by almost eighty percent.

What do we lose when we stop needing each other?

The decline in collaboration by eighty percent is not just a statistic. That’s eighty percent fewer conversations over coffee about how someone else solved a similar problem. Eighty percent fewer situations when a younger programmer asks an older one why they chose this approach and not another. Eighty percent fewer moments when a random exchange of ideas gives birth to something that no one would have come up with on their own.

The study shows a paradox that is particularly pronounced in the case of juniors. They are the ones who benefit the most from AI: twenty-two percent more contact with new programming languages, faster learning, lower barrier to entry. And at the same time, it is juniors who are the first candidates for reduction in companies that see AI as a way to cut costs. It’s like cutting down seedlings because adult trees give more shade.

But there is something even deeper. Knowledge in organizations does not exist only in documentation and code. It exists in relationships, in shared experiences, in the thousands of small interactions that build what we call organizational culture. As AI takes on the role of “colleague to ask,” this fabric of relationships begins to thin out. Is collaboration with other people just an obligation resulting from the breadth of tasks in projects? Or is it something else: a source of meaning, belonging, a communal understanding of why we do what we do? A conversation that does not lead to an immediate solution to a problem can build trust that will pay off in a year. A meeting that feels like a waste of time may be the only place where people feel part of something bigger.

The strategic implications are clear: replacing juniors with AI is short-sighted. AI should support their development, not eliminate them from the organization. But the human implications are harder to measure. We are entering an era where we can be productive alone. The question is whether we want to. And whether we are aware of the price we pay by choosing efficiency at the expense of community.

This post is part of the project “People and Algorithms in Organisations: Competences to Work in the Digital Environment”, funded by the NAWA – Narodowa Agencja Wymiany Akademickiej (Polish National Agency for Academic Exchange).

#AI #ArtificialIntelligence #genAI #collaboration #FutureOfWork #ResponsibleAI #Productivity  #DIGIT #UEP #PUEB