The Architecture of How We Actually Work
I recently attended a talk "From Solo Performances to Collective Brilliance," hosted by User Stories and given by Mireille Pellegrini Petit. It was a talk about teams. What really stayed with me from this session was the idea of psychological safety and how critical it is for collective success.
Pellegrini Petit opened with something that challenges a lot of how we think about building strong teams. The most talented team you've ever seen and the most effective team you've ever seen are rarely the same team. The instinct is to assume that if you gather enough skilled, experienced people in one room, a high performing team is a natural consequence. It isn’t. When talented teams underperform, the reason is almost never a lack of intelligence or skill. It's what happens between the people: ego, competition, silos, the quiet fear of being wrong in front of someone senior. None of that shows up on a CV.
She called this relational architecture, and drew a direct comparison to the kinds of architecture organisations are far more comfortable investing in and talking about. Technical architecture. Product architecture. Data architecture. We plan those meticulously, argue over them in design reviews, revisit them when they stop serving us. We almost never plan the architecture of how people actually relate to one another while doing the work, even though it's the layer most likely to determine whether any of the other architectures get used well. It's also the layer most likely to be neglected precisely because it's the hardest to point to, and the easiest to notice only once it's already broken.
During the talk she spoke about Google's Project Aristotle, the two year study of 180 teams that tried to isolate what actually predicted performance. Seniority, composition, experience, all of it mattered less than expected. The strongest factor by a distance was psychological safety. Psychological safety isn't a team where everyone gets along and nobody says anything difficult. It's a team where people feel safe enough to say the difficult thing, to admit a mistake or flag a concern without expecting to be humiliated or punished for it. Conflict isn't the enemy of a psychologically safe team. Silence dressed up as agreement is.
One thing that also caught my attention was a slide on social pain. The brain regions activated by exclusion or dismissal overlap with the regions activated by physical pain. Being talked over in a meeting and being physically hurt aren't as different, neurologically, as we like to pretend. Relational architecture, in other words, isn't a metaphor. It is rooted in physiology.
The projects where I've done my best work, made my clearest judgment calls, caught a problem before it became expensive, were never the projects with the smartest people in the room. They were the ones where I trusted that saying "I think this is wrong" wouldn't cost me anything. The projects where I phrased the concern three different softer ways before finally not saying it at all, were never short on talent either. So the question I keep asking myself isn't whether a team is smart enough. It's whether it's safe enough for the smart parts to actually reach the surface. And this matters more now, not less, with AI in the mix. A tool can only work with what a team is willing to say out loud.