The standard story about psychological safety runs through the relationship between a worker and their boss. Can the worker speak up? Will the boss punish them? Will the team learn from a mistake or pin it on the person who made it? These are the dynamics the literature focuses on, and rightly so. They cover most of where the high-leverage work happens.

But there's a version of the same problem that operates sideways, between coworkers at the same level, and it gets a lot less attention. I came across a clean illustration of it in a webinar with Nick Shonsky, Director of Continuous Improvement at The Standard Group, a printing company in Pennsylvania.

The Standard Group was built from eight acquired companies over several decades. Each company brought its own culture, vocabulary, and ways of working. As they built out their Lean program, they ran a survey asking employees what they saw as the biggest obstacles in their work. One of the patterns that came back: their production tickets were a mess.

The tickets are the artifact that passes between two groups. Production planners write them. Production operators read them and use them to set up jobs. If the ticket is accurate and complete, the operator can do the work efficiently. If the ticket is wrong, the operator either runs the job badly or has to track down the planner to clarify.

Here's the dysfunction Nick describes. The operators stopped reading the tickets carefully because they assumed the tickets would be wrong. The planners stopped investing in making the tickets right because they suspected the operators wouldn't read them carefully anyway. Each side's behavior confirmed the other side's assumption. The dysfunction was stable. Both sides were right about the past, and neither side could fix it alone.

This isn't a story about hostile coworkers. It's not about someone punishing a peer for speaking up, which is the version of horizontal psychological safety that gets the most attention. There's no villain here. There are two groups of workers, both behaving rationally given what they'd learned about the other group over years of working together, and the rational behavior was producing collective dysfunction.

The standard interventions don't work on this. A trust workshop doesn't address it -- the trust isn't missing because people don't know how to trust, it's missing because the artifact running between them keeps reinforcing the reasons not to. Getting both sides in a room together to talk it through doesn't address it either -- the conversation might be productive, but the next morning the tickets are still the same tickets, and the same behavior produces the same result. Coaching either group on communication misses the point entirely. The communication channel itself is the problem.

What actually worked at The Standard Group was a redesign of the tickets. Standardized terminology -- same fold called the same thing regardless of which acquired company the planner originally came from. Consistent layout, so the operator knew where to find the information they needed. Accurate content, because the planners committed to making it accurate. And then the operators committed to reading them carefully.

The first cycle was tentative on both sides. The planners produced a careful ticket, half-expecting the operator to skim it the way they always had. The operator read it carefully, half-expecting to find errors that proved their old assumption right. The second cycle was less tentative. The third was less still. After a few months, the trust that hadn't been there at the start was operating as a normal feature of the workflow.

The principle worth pulling out: the artifact change made the new behavior possible. The behavior held over time, which produced the trust. The trust didn't come first.

This inverts an assumption most of us carry. We tend to think trust has to be present before cooperation can happen -- that the workshop, the team-building, the relationship work has to come first, and the operational improvements follow. Sometimes that's right. But sometimes the trust gap is being held in place by an artifact that's making rational cooperation impossible, and no amount of relational work will produce sustainable change as long as the artifact stays the same.

You see this pattern in a lot of places once you start looking for it. Engineering and sales who don't trust each other, with requirements documents as the artifact in the middle. Clinical and billing teams in a hospital, with documentation as the artifact. Development and QA, with bug reports. Day shift and night shift in any operation, with the handoff log. Each pair of functions has its own version of the production ticket -- some artifact that's supposed to carry the trust and instead carries the dysfunction.

A few practical implications for leaders running improvement work in any of these environments.

The first is diagnostic. When two functions don't trust each other, look at what passes between them before you look at the people. The artifact often tells you what's wrong faster than the conversations will. If the artifact is unreliable, the distrust is rational, and no amount of coaching will fix it.

The second is sequencing. Fix the artifact first. The relational repair is real work and worth doing, but it's downstream of the structural change. If you try to do it the other way -- workshops first, artifact change later -- the workshops produce optimism that erodes the next time someone reads a bad ticket. Trust built on top of bad infrastructure is fragile.

The third is patience. The first few cycles after the artifact changes will be tentative on both sides. People who have been burned for years don't trust new behavior immediately. They need to see it hold, multiple times, before the old assumptions begin to update. The work is slower than the workshop version. It's also more durable.

Trust doesn't come first. It comes after the conditions that make trust rational are in place. The work of building those conditions is structural before it's relational, and naming that distinction is what makes horizontal psychological safety addressable rather than just nameable.