A document sent for comment lets every reviewer register a small objection without anyone having to reconcile them, the result is a compromise nobody chose. A working session forces the same disagreements into a room where they have to be resolved live, in front of the person who owns the decision.
A delegate cannot make the call a working session is built to force, which is why the founder’s presence is treated as the product, not a scheduling inconvenience.
What a circulated document actually optimises for
It is worth being precise about why the document-for-comment pattern fails, because the failure is not laziness or bad process, it is a predictable consequence of the format’s own incentive structure. A document sent for asynchronous review invites every stakeholder to register an objection cheaply, with no obligation to reconcile it against anyone else’s objection, and no forcing function that requires the group to actually agree on a single direction before the process ends. Each reviewer optimises locally, for "did I flag my concern," rather than globally, for "did the group reach the right answer", and a document that has absorbed a dozen locally-reasonable, mutually-uncoordinated edits usually ends up as a compromise nobody on the team would have chosen if they had been forced to pick one direction and defend it out loud.
This is not a hypothetical failure mode, it is close to the default outcome of any comment-based review process run over enough cycles, because the format structurally rewards hedging over commitment. A reviewer who raises a sharp objection risks being the person who blocked consensus. A reviewer who registers a soft caveat and lets the document proceed risks nothing. Over several rounds of revision, the soft caveats accumulate and the sharp objections get diluted into footnotes, and the resulting document reflects the path of least resistance through a dozen individually reasonable but collectively incoherent edits.
A session runs in four parts: framing states the specific decision on the table as a question with a forced choice, evidence is argued rather than just presented, the decision-maker chooses live on the call, and the decision and its reasoning get written up the same day rather than reconstructed later.
Why the four parts have to run in that order
Framing is the part most sessions get wrong when they fail, because a badly framed decision produces a badly forced choice even if the rest of the format runs perfectly. "What do you think about the pricing page" is not a framing, it is an invitation to the same diffuse, uncommitted feedback a circulated document would have produced anyway. "Should the pricing page show four tiers or two" is a framing, because it names the actual fork in the road and makes it impossible to answer without picking a side. Most of the discipline in running this format well lives in getting the framing this specific before the call even starts, not in anything that happens during the call itself.
Evidence gets argued, not just presented, because the point of the session is not to inform the decision-maker, a document could do that, it is to stress-test each option against the strongest version of the case for the other one, live, with both sides in the room able to respond to a challenge in real time. A document-based process lets a weak argument sit unchallenged if nobody happens to raise the specific objection in writing. A live argument forces the weak points to surface in real time, because the other side is sitting there, motivated to find them.
The decision-maker choosing live, on the call, rather than "taking it away to think about it," is the step that actually converts a good discussion into a real decision. A choice deferred to private reflection after the room disperses quietly reopens every option the session just closed, the decision-maker, alone, without the counter-arguments freshly live in the room, tends to drift back toward whichever option felt most comfortable before the session started, undoing the actual work the live argument just did.
Same-day write-up matters for a more mundane but equally real reason: memory of a decision’s actual reasoning degrades fast, and a decision reconstructed a week later from someone’s recollection of the call routinely differs from what was actually decided, in ways nobody notices until two team members act on two different versions of the same meeting. Writing the decision and its reasoning down within hours, while the argument that produced it is still fresh, is cheap insurance against a specific, common and entirely avoidable failure mode.
What it requires from the client is that the owner actually shows up. A session run without the real decision-maker in the room produces a recommendation, not a decision, and a recommendation is not what the format is built to output.
Why a delegate genuinely cannot substitute
This constraint is worth defending rather than treating as a scheduling preference, because the instinct to send a senior deputy instead of the actual founder is reasonable on its face and wrong on the mechanics. A delegate, however capable, is structurally unable to make the specific kind of call a working session is built to force, not because delegates lack judgement, but because they lack the authority to bind the organisation to a decision without checking back with the person who actually owns it. Every choice a delegate makes in the room carries an implicit "pending confirmation," which quietly reopens the same deferred-decision problem the live format exists to close.
A session that runs with a delegate in the founder’s seat produces something that looks identical to a real session from the outside, the framing happens, the argument happens, someone says a decision out loud, and is functionally a very well-run recommendation, because the actual owner still has to review it afterward, potentially disagree with it, and the whole cycle of async back-and-forth the format was built to eliminate quietly reappears one level downstream. The founder’s presence is not a courtesy to the format. It is the specific ingredient the format cannot function without, because authority to decide is not delegable in the same way expertise to advise is.
What good framing looks like in practice
A useful test for whether a proposed session topic is actually ready to run: can the framing be written as a single sentence with exactly two or three named options, none of which is secretly "let’s discuss further." Naming which foundation a venture is actually missing before scheduling any session at all is often the framing work that has to happen first, a session scheduled before anyone has done that diagnosis tends to produce a vague, unforced discussion, because there is no sharp enough question yet for the format to force an answer to.
Topics that resist this kind of framing are usually a sign the actual decision has not been isolated yet, not a sign the format does not apply. "How should we think about our brand" is not ready for a session. "Should the homepage lead with the founder’s personal story or the product’s mechanism" is. The work of narrowing the first into the second is real, upstream work, and skipping it is the most common reason a working session produces a mediocre result, not because the format failed, but because it was pointed at a question too diffuse for any format to force a clean answer to.
What this format is not built to do
It is worth being equally clear about the format’s limits, because the temptation once a team has seen this work well is to reach for it on every decision, including ones it is genuinely the wrong tool for. A working session is built to force a choice between a small number of already-identified options, it is not built to generate those options in the first place, and running a session to "brainstorm directions" tends to produce the same diffuse, unforced discussion the format exists to prevent, because there is no sharp fork yet for the live argument to test.
The upstream work, narrowing an open-ended question down to two or three genuinely competing, well-understood options, is a different kind of work, done differently, usually by fewer people over more time, before a session is ever scheduled. Confusing the two, and expecting a live session to do both the narrowing and the deciding in one sitting, is the reliable way to burn the format’s credibility on a decision it was never built to handle.
What the write-up is actually for
The same-day write-up deserves more attention than it usually gets, because it is doing more than record-keeping. A decision written up while the reasoning is fresh becomes the shared reference every future disagreement gets checked against, not "what did we decide," which memory answers unreliably, but "what did we decide and specifically why," which is the part that actually prevents the same argument from happening again in three months once someone has forgotten the counter-argument that closed it the first time.
A good write-up names the option that was rejected and the specific reason it lost, not just the option that won. This matters more than it sounds like it should, because the rejected option almost always resurfaces eventually, a new hire proposes it fresh, not knowing it was already argued and closed, or the original decision-maker, months later and under different pressure, starts drifting back toward it. A write-up that only records the winning choice gives that resurfacing nothing to push against. A write-up that records why the alternative lost gives the team a fast, specific answer the second time the question comes up, rather than re-running the entire argument from scratch.
Why this format resists documents-as-comfort
There is a specific, uncomfortable reason teams default back to circulated documents even after experiencing a working session that clearly worked better: a document that has quietly absorbed everyone’s soft objections feels safer in the moment than a decision that one person made out loud, live, in front of the room. Nobody personally owns the outcome of a compromise nobody chose, if it goes wrong, the diffuse responsibility spreads thin enough that no individual bears much of it. A live decision has a name attached to it, and that visibility is exactly what makes the format effective and exactly what makes it uncomfortable to a culture more used to distributing risk than concentrating accountability.
That discomfort is not a flaw to be engineered away. It is the mechanism working as intended, a decision that visibly belongs to someone tends to get made more carefully than one nobody can be held to, precisely because the person making it knows their name is on it. Softening that visibility to make the format more comfortable would remove the exact property that makes it produce better decisions than the alternative it replaced.
The format at scale
A natural question once a team has seen a handful of working sessions run well: does this scale past a small founding team to a larger organisation with more stakeholders and more layers of authority. The honest answer is that the format scales by staying narrow, not by growing to include more people in the room. Adding stakeholders to a session does not make the decision more representative, it recreates the diffuse-objection problem the format exists to solve, just live instead of asynchronous, with a dozen people each registering a spoken caveat instead of a written one.
The version that scales correctly keeps the room small, the actual decision-maker, plus whoever is bringing the strongest case for each option, and pushes broader input to before the session, during the framing stage, rather than into the room itself. By the time the live session runs, the options should already be well-understood by everyone attending; the session’s job is to force the choice between them, not to educate the room on what the choice even is. A larger organisation running this format well has more sessions, each still small, rather than fewer sessions with more people crowded into each one.
