The Infrastructure of Belonging: Which Teams Will Survive AI?

During COVID we proved a strange thing: you can keep a company running through a grid of faces - and still watch the human glue quietly dissolve. Hour after hour of 'being on' in front of a camera didn't just tire us out; it thinned us out.

Psychologists even gave it a name. 'Zoom fatigue' - arguing that video calls overload us with sustained eye contact, self-monitoring, and missing body-language cues. Stanford's Jeremy Bailenson has written extensively about this. And large-scale workplace data backs up what it felt like: when work goes fully mediated, our networks tend to narrow, weaker ties fade, and silos harden - meaning fewer of the small interactions that make a team feel like a team. Microsoft researchers have published on this shift in collaboration patterns.

So we adapted the only way we could: we doubled down on knowing the people behind the pixels. I’m sure most of you did the same as me. More check-ins, more context, more 'how are you really?' Because without deliberate presence, relationships decay. That lesson matters now, because AI will make it even easier to operate without humans ever fully meeting - and belonging doesn't survive on efficiency; it survives on an infrastructure of belonging.

The Stakes of the Great Re-bundling

This blog is about why the institutions that survive the AI age won't be the most efficient. They'll be the ones that understood what efficiency destroys.

Why does this matter? Because we're at an inflection point. Remote workers consistently report higher rates of loneliness - Gallup found 25% of fully remote employees experience it compared to 16% of on-site workers. The majority of young people now report anxiety around phone use. A landmark meta-analysis found empathy among college students dropped 40% between 1979 and 2009, though recent data suggests this trend may be reversing. These are all consequences of the same underlying shift.

Unfortunately, most organisations respond to connection problems by buying better collaboration software. They optimise for throughput when the actual problem is trust. They schedule more video calls when the issue is that nobody has shared a meal in eighteen months.

If you can't tell the difference between transactional collaboration and relational collaboration, you'll wake up one morning with a well-documented team that falls apart under pressure. You'll have children who send AI-drafted birthday messages to grandparents they barely know. You'll have 'communities' that exist only as notification preferences.

Now my strong hunch here is that the friction we may be tempted to remove might be the point. The inefficiency is a feature, not a bug.

I'm going to show you how I think about what I call "the infrastructure of belonging" before AI completes the unbundling.

Given my focus on humanity-first when it comes to AI, you may forgive me for thinking this is the most important blog of the Big Questions series.

With that, let's dive in.

The Great Unbundling: What We Traded Away

Twenty years ago, we ate dinner as families because there wasn't an alternative. We asked neighbours for recommendations because there was no algorithm. We called our parents because texting didn't exist. These weren't explicit choices, but rather they were constraints. And within those constraints, relationships accumulated.

Technology spent two decades systematically removing those constraints. Family meals gave way to food delivery apps (now a $1.23 trillion market serving 60% of consumers weekly). Neighbourhood knowledge migrated to Nextdoor. Elder wisdom was replaced by Google. Workplace camaraderie dissolved into Slack threads and async video updates.

Each unbundling solved a genuine friction. Each also severed a relational thread. We gained convenience.

But in many cases we lost the people who show up when we're unimpressive at best, or in genuine trouble at worst.

The cost is finally visible at a societal level. Clearly beyond the scope of this blog, but (and perhaps stating the obvious) I have a hunch that the societal fragmentation we see daily in the media is a mirror for this underlying shift.

The Collaboration Paradox: Why Teams Need Bodies, Not Better Tools

I've spent thirty years watching organisations try to solve human problems with technology. The pattern is consistent: we assume that better tools create better collaboration. The research says otherwise.

Video calls can't replicate the neurochemistry of presence. Research on trust in hybrid teams shows that remote work "reduces opportunities for informal interactions and shared identity formation, leading to communication gaps and trust deficits." Trust builds at rates that screens cannot accelerate. The weak ties, the spontaneous corridor conversations, the lunch that runs long because someone finally admits they're struggling, these don't happen on scheduled video calls. I vividly remember my first day back in the office after the initial covid lockdown – genuinely excited about wearing “nice” shoes again and the joy at bumping into people in the office was palpable (and I’m sure people thought I was slightly mad at the shoe thing).

This is the core problem. AI cannot build the relational “substrate” that lets teams survive conflict, ambiguity, and change. It can summarise your standups perfectly while your team quietly falls apart.

The practical insight? Budget for quarterly gatherings, not better software. As one leadership study noted, managers now "must consciously plan to meet with employees" and "be much more sensitive to notice which moods, fluctuations or topics are currently developing. It's easier in person." The ROI of in-person time is invisible until you need it. Then it's everything.

The Rebundling: Protecting What Can't Be Replicated

Families, close friendships, committed teams, these are "last-mile relationships." AI can't replicate them because their value lies in the unconditional presence. The grandmother who listens without optimising. The colleague who notices you've gone quiet. The friend who shows up when there's no reward for showing up.

The institutions that thrive in the AI age will be those that deliberately protect inefficiency. The dinner that runs long because someone needed to talk. The team offsite where sessions are held without an agenda. The community gathering with no purpose but showing up. These aren't bugs to be fixed. They're features that create the trust reserves we'll need when things get hard.

This isn't anti-technology. It's recognising that some things exist because of the effort and presence, not despite them. The synthesis is practical:

- Use AI for the transactional: scheduling, reminders, documentation, meeting prep, follow-up tracking.

- Protect the relational from optimisation: rituals, unstructured time, physical presence, the conversations that go nowhere.

- Recognise that the friction you're tempted to remove might be the point. The inefficiency might be building exactly what you'll need when efficiency fails you.

The Practice of Presence

The organisations getting this right share a common trait: they've stopped treating presence as a scheduling problem and started treating it as a design constraint.

This shows up in unexpected ways. The most resilient teams share one characteristic: they've accumulated what sociologists call 'interaction residue' - the accumulated small moments that create shared context. Inside jokes. Collective memories of projects that nearly failed. The kind of mutual knowledge that can't be onboarded.

The physical dimension matters more than most remote-work advocates admit. Not because offices are inherently superior, but because certain types of trust seem to require what anthropologists term 'co-presence'.

Perhaps most counterintuitively, the teams that sustain belonging over time are the ones that protect unproductive time. Not team-building exercises with forced fun, but genuine slack in the system - conversations that go nowhere in particular, relationships maintained without agenda. This is expensive. It looks like waste on a utilisation dashboard. But it's also the substrate on which trust actually grows.

The Bottom Line

We spent two decades unbundling social institutions in pursuit of convenience. AI is completing that project, handling the transactional parts of human connection with unprecedented efficiency. The old framing was that friction was the enemy, that optimising away inefficiency would give us better relationships with less effort.

The new paradigm recognises that some relationships exist because of the friction, not despite it. The value is in the commitment, the presence, the willingness to show up when leaving would be easier.

I increasingly think the institutions that survive the AI age won't pursue efficiency to its ultimate conclusion. They'll be the ones that understood what efficiency destroys. Family. Team. Community. Friendship. These aren't inefficiencies to be optimised. They're the last defensible moat.

AI can mediate, summarise, translate, and remember. What it cannot do is show up when there's no reward for showing up. That's not a limitation of the technology. It's a definition of what makes us human.

Until next time, you'll find me lingering over dinner, answering the phone, and budgeting for the meetings that have no agenda.