The two teams that have never spoken
August 22, 2026
They report to the same VP. They serve the same customer. On the org chart they sit side by side. In the record of who actually talks to whom, there is nothing between them at all.
The question that took four days
A customer asks whether a particular configuration is supported. The account manager does not know. Somebody in engineering knows. The answer is settled, uncontroversial, and about ninety seconds long. Getting to that ninety seconds takes four days.
It goes like this. The account manager asks their own team, because that is who they know. Someone suggests a name. That person has left, or has moved teams, or answers with a caveat that is really a redirection. A meeting is proposed. Eventually a colleague of a colleague forwards it to the right engineer, who replies in a sentence and is mildly puzzled that it took so long. Everyone behaved well. Nobody was unhelpful. The customer waited four days for a fact the company has known for a year.
This is the elephant we call Silo: information trapped inside a team boundary. It is not caused by anyone withholding anything. It is caused by the boundary being the only structure that reliably exists.
The evidence
A quarter of the week, spent looking
25%
of teams' time goes to just searching for answers
Atlassian, State of Teams 2025 (n=12,000 knowledge workers and 200 executives).
Isn't that just a search problem?
It looks like one, and the distinction is the whole point of this elephant. Searching a space you belong to is a tooling problem: the thing is somewhere, the index is bad, better software helps. Searching across a boundary is a social problem, and better software does very little for it.
The difference is what you are actually missing. Inside your own team you know roughly who would know, and the search is for the artifact. Across a boundary you do not know who would know, and finding out means spending something: a favour, an interruption, the small exposure of admitting you do not know. People are rational about that cost. They ask the person they already talk to, get a partial answer, and proceed on it.
Which is why the twenty-five per cent is not spread evenly across the week. It concentrates at the seams. The same question asked inside a team takes a minute. Asked across two of them it takes four days, and the four days are almost entirely spent finding a human, not finding a document.
What does a silo actually cost?
The waiting is the visible part and the least of it. Underneath sits local optimization: each team, rationally, improving the metric it is measured on, using the picture it can see. Support optimizes for closing tickets, so it stops escalating the awkward ones. Engineering optimizes for shipping, so it never learns which awkward ones recur. Both teams look good. The product keeps the defect.
Then there is the version of the truth problem. Two teams answering the same customer from different partial pictures will eventually give that customer two answers, and the customer, unlike anyone internally, sees both. This is the point at which a silo stops being an efficiency matter and becomes a credibility one.
And the compounding cost: a boundary that is expensive to cross gets crossed less, which makes it more expensive to cross next time, because the relationships that would have made it cheap never form. Silos are self-reinforcing. That is what makes them worth deliberate attention rather than goodwill.
How do you know the Silo is in your room?
Nobody experiences a silo as a silo. From inside it just feels like a normal week. Look for what routes around it.
- One or two people are on every cross-team thread, because they are the only route between groups.
- A question's answer time depends on which team is asking, not on how hard the question is.
- Teams have different names for the same customer, product or stage, and translation happens in meetings.
- Onboarding takes far longer than the work warrants, because the real map is a set of relationships nobody wrote down.
- A decision gets remade because the team making it did not know it had already been settled elsewhere.
Can you actually see a silo?
Yes. And this is the part most organizations do not realise is available to them. A silo is not a mood or a culture complaint. It is a structural property of who works with whom, and collaboration leaves a trace: who is on which threads, which groups exchange work, where requests stop.
Organizational network analysis reads those patterns and draws the result. What appears is not a re-drawn org chart but the network that actually operates underneath it, and silos show up as exactly what they are: a gap where a connection should be, with a small number of people carrying every crossing. Those people are usually the first thing the picture reveals, because they are simultaneously the organization's most valuable connectors and its most fragile single points of failure.
This matters because it converts an argument into a measurement. “Marketing and Delivery do not talk” is a claim someone will dispute in a meeting. A picture showing two dense clusters joined by one person is not disputable, and it also tells you precisely where to intervene, which is rarely where the reorganisation would have put its energy.
How do you get the Silo out of the room?
The same three moves we bring to any elephant, pointed at this one.
Map
Measure the network before changing it. A few weeks of analysis shows where work actually crosses, who carries the crossings, and which boundary is costing the most, as opposed to which one is complained about most. These are frequently not the same boundary.
Prove
Pick the single most expensive seam and make one crossing cheap: a standing point of contact, a shared definition, a route that does not depend on knowing a name. Then measure the same question's answer time again. One seam, one cycle. A connection you can demonstrate beats a reorganisation you have to argue for.
Scale
Extend what shortened the path, and keep watching the structure rather than the sentiment. The goal is not a flat organization. Boundaries are how work gets divided. The goal is that crossing one costs a message instead of four days, and that no single person is the only bridge.
Find out where your organization stops talking to itself
The Elephant Safari names your herd in ten questions and ranks them by what they cost you, and sketches your network as it goes. Or start from the business problem instead: systems and teams that do not join up. Get the Silo out of the room. Make one crossing cheap.
Every figure in this article traces to a primary source. See it in Knowledge