We all know a “Dave.” Dave is the senior engineer who knows exactly why the legacy database does that weird thing at 3:00 AM on Tuesdays. Three years ago, if you needed to understand that database, you had to go talk to Dave.

You had to send a Slack message. You had to endure Dave’s mandatory five-minute rant about the original architecture. You had to experience a necessary moment of professional vulnerability by admitting, “Hey, I don’t understand how this works.”

Today? You just feed the repository into an AI model and ask it to explain the schema. You get your answer in 12 seconds. No rants. No social friction. No waiting for Dave to finish his coffee. It is peak efficiency.

It is also the quiet death of your team dynamics.

The core thesis of this emerging research is both painfully obvious and easily ignored: When we replace human colleagues with AI assistants, we optimize for speed but bankrupt our social capital.

The tech industry has become obsessed with “frictionless” workflows. We treat every human interaction as a bottleneck to be engineered out of the system. AI is the ultimate frictionless colleague. It never judges you for asking a stupid question. It doesn’t care if you forgot the syntax for a basic SQL join for the fifth time this week. It just gives you the answer, politely and instantly.

But here is the psychological reality we are forgetting: Trust is built in the friction.

When you ask a colleague for help, you are engaging in a micro-transaction of vulnerability. You are saying, “I need you.” When they help you, they are saying, “I have your back.” That slightly awkward, time-consuming exchange is the mortar that holds the bricks of a team together.

If you remove all the mortar because it takes too long to dry, you don’t get a stronger wall. You get a pile of loose bricks waiting to be knocked over.

What happens when a team of engineers, marketers, or product managers spends 80% of their day collaborating with an LLM instead of each other?

You create a room full of brilliant, highly productive silos.

On paper, velocity goes up. Jira tickets are burned down at record speeds. Management is thrilled. But the underlying fabric of the team begins to rot. When a catastrophic, unscripted failure happens – when the servers melt down, a major client threatens to leave, or a massive strategic pivot is required – you suddenly need high-trust, high-bandwidth human collaboration.

But your team hasn’t practiced that muscle in months. They’ve spent the last year interacting with an obsequious chatbot that is hardcoded to agree with them. They have forgotten how to debate, how to compromise, and how to navigate the messy, beautifully complex human ego. They try to solve human crises with the same isolated, transactional mindset they use with their AI. And it fails spectacularly.

We cannot – and should not – put the AI genie back in the bottle. I am not advocating for a return to the dark ages of waiting three days for an email reply to a question that a machine could answer in three seconds.

But we must recognize that if AI is going to handle the technical heavy lifting, we have to deliberately engineer human connection back into our days. We have to make empathy and collaboration a choice, rather than a byproduct of needing information.

This means occasionally pinging Dave to ask his opinion on an architecture decision, even if Claude could give you a decent pros-and-cons list. It means holding design reviews not just to catch bugs, but to practice the art of giving and receiving human feedback. It means recognizing that the goal of a company isn’t just to generate output; it is to cultivate an environment where human beings actually want to spend their time.

The machines have freed us from being mere typists. Don’t immediately squander that freedom by turning yourself into a hermit. Go bother your colleagues.

More in: https://hbr.org/2026/10/team-dynamics-weaken-when-employees-turn-to-ai-instead-of-colleagues

The choice, is as per usual – yours!