The Boule

How do you run a Spec Session across more than one team?

The Boule was a council of five hundred, fifty chosen from each of ten tribes. It didn’t replace the Assembly. It prepared what the Assembly would decide, and carried the city’s business in the space between full gatherings.

Trireme answered a narrower question than it looked like it was answering. Three functions, one team, agent as the fourth participant. Fine for a room small enough to fit around one spec. But most engineering organizations aren’t one team. They’re several, each with its own Agora, each producing its own shared understanding, and none of that automatically holds together at the seams. I don’t have this figured out. Nobody’s run it long enough to know what breaks. This is a theory, not a report from the field.

The lazy answer is a bigger room. Take the Agora, invite every team, run one enormous Spec Session. That’s the Ekklesia, not the Boule, and the Ekklesia is exactly the failure mode Nobody’s Walking Over to a Desk already named. Too many people, no real decision, everyone nodding toward a consensus that isn’t real. Scale the room and you don’t get more shared understanding. You get a meeting that can’t end. LeSS was built on the idea that scaling shouldn’t mean adding layers on top of Scrum, it means figuring out how to apply the same principles at a larger scale, as simply as possible. I respect the instinct. SAFe took the opposite route, adding program and portfolio layers, release trains, and formal planning events to coordinate many teams at once. Both exist because someone already tried to solve exactly this problem with structure. I don’t think either is the answer here. Neither was solving the problem underneath the problem.

Room size isn’t the constraint. Ownership is. Who fixes the bug when three teams touched the feature. Who’s accountable when the spec that shipped was agreed on by people who don’t sit in any of the affected teams. Who’s on the Hook already argued the team is the unit of accountability, not the individual. Cross-team work asks the same question one level up, and doesn’t have as clean an answer yet. Here’s the shape I keep landing on, provisionally. Each team sends representatives to the cross-team session. Two, not one. A single rep is a single point of failure for what their team actually meant, the same risk a spec carries when only one person wrote it. Two gives the room a chance to surface disagreement the rep’s own team hadn’t fully resolved yet, before it gets exported as a decision on that team’s behalf. A stakeholder manager, likely a PM, gives direction: what’s actually in scope for this cross-team decision, what isn’t. That’s the Boule. Not the whole team, not one person speaking for everyone. The original Boule drew fifty representatives from each of ten tribes, distributed roughly by population, so every part of the city had a proportional voice without needing the whole citizenry in the room. Its central job wasn’t to decide policy itself. It drafted the matters that would go before the full Assembly, and the Assembly voted on what the Boule had already shaped. That distinction is the one worth keeping. The Boule produces alignment, not accountability. Accountability still lives where it always did, in the team that has to carry what shipped. Which means a session isn’t done when the room agrees. It’s done when what the room agreed on can be handed back to each team as something that team can actually own, not a directive arriving from outside, but a commitment they can be on the hook for. If a cross-team decision can’t decompose back into team-owned work, the session didn’t finish its job. It just moved the ambiguity somewhere less visible.

What it produces still has to get back to the people who weren’t in the room. That’s not a Boule problem. It’s a different mechanism entirely. Call it decision formation versus decision propagation. The Boule is formation, a smaller room producing the same currency a single-team Spec Session produces, shared understanding, just among representatives. Propagation is what happens after. How that understanding reaches everyone who wasn’t there. The Amphictyony model, independent governments, each keeping full autonomy, cooperating only where they share something specific, might be closer to how propagation should work than anything resembling a hierarchy. Whether it needs its own institution, or can grow organically with a gate that checks it actually landed, is the next question. It’s its own post, not a paragraph inside this one.

I keep coming back to the same tension. Fast, but with shared understanding intact. Less back-and-forth between teams, not more process wrapping the back-and-forth in ceremony. The Boule doesn’t resolve that tension. It’s a shape that might hold it better than a bigger Agora would, and better than borrowing a scaling framework built for a different problem. If this fails, it won’t be because the room was the wrong size. It will be because ownership never made it back to the teams that had to carry it.

References