The Connection Deficit Audit.
A five-minute self-check on the moments your team used to learn from, and whether they still happen.
From Episode 1 of the Average Robot podcast with Natalie Emirzian, who spent 15 years leading HR and transformation.
Most of us learned the job by being around people who already knew it. You'd ask a colleague how to do something, and while they showed you, you'd pick up how the place really worked. A lot of those questions go to AI now, and for plenty of them that's the right call. It's the other questions that matter - the ones where you were learning how to read a room or work out why a decision got made the way it did. This checks where those moments still happen on your team.
- 1
The questions that stopped getting asked out loud
Asking a person a basic question has always carried a small cost - you might look like you should already know. Asking AI carries none, so the small questions go there by default now. That also means the low-stakes back-and-forth that used to build working relationships happens less often.
When did someone last ask a colleague a simple question out loud, instead of a tool?
- 2
The work that used to train your juniors
Summarising a report, cleaning a dataset, drafting the first version. Nobody misses that work, but it's how new people built judgment - and it's exactly what AI does well. Stanford's Digital Economy Lab found roughly a 13% relative drop in employment for workers aged 22 to 25 in the jobs most exposed to AI, while older workers held steady.
Are your newest people still getting the reps that build judgment, or is that work going to the machine?
- 3
What people pick up just by being near each other
A lot of learning is someone glancing at your work and saying "not like that, like this" before you've gone too far. An NBER study of software engineers found that sitting near teammates increased the feedback people got on their work by 18.3%, and the younger ones gained the most. When the quick questions route to AI and the desks empty out, that everyday contact gets thinner.
Are your people still close enough to each other to learn in the moment?
- 4
The things that were never written down
Natalie's list on the podcast: how to navigate organizational politics, influence stakeholders, manage conflict, understand why a past decision got made the way it did. You don't learn those from a document, because nobody wrote them down. They pass between people close enough to watch how it's actually done.
Are your people picking up the unwritten stuff, or only the written stuff?
- 5
The standing you build just by being around
When people work near each other, they build a reputation with each other without trying. Someone sees how you handle a bad week, learns they can count on you, and later puts your name forward in a conversation you're not part of. Natalie put it directly: "You need people to vouch for you, career sponsors. You're not going to get that without those interactions." You can't ask a chatbot to vouch for you.
Is your team spending enough real time together to build that kind of standing?
- 6
Drowning in the tools
Natalie described being able to disappear into the tools and not speak to anyone all day. It's a real pull, and it has a cost. Gallup reports about one in five employees worldwide feel lonely, and that disengagement drains something like 9% of global GDP. A team that's individually productive but disconnected can look fine on a dashboard for a while.
Do your people feel part of something, or just efficient and on their own?
Your read.
Mark all six above to see your read.
(0 of 6 marked)
Three ways to protect it on purpose.
- 1.Put the accidental learning back on purpose - office hours, shadowing, an "ask a person first" habit for the questions worth it.
- 2.Protect the time people spend actually working together, and hold the hours for it, or it's the first thing cut when things get busy.
- 3.Decide what the freed-up time is for. Some of the hours AI hands back should go to the human side by design.
Where to take this next.
This looks at the connection between people at work. There's a companion self-check, the free AI Alignment Snapshot, that measures a different thing: whether your business intent, your actual AI activity, and your customer value line up. Two honest reads for the same leader.
Average Robot - change management for AI at work.