Conversation Cafe: What to Do With a Decision the Team Didn't Get to Make

Leave Balance had been sold on flat per-company tiers since day one — under fifty employees, under two hundred, unlimited. In June the decision came down to move to per-employee pricing at renewal, and it was not a decision the product team was consulted on. The unit economics were genuinely broken at the top tier, where a handful of two-thousand-person customers paid the same as a two-hundred-person one and generated forty times the support load. I agreed with the change. That didn’t make the Monday easier.

The people who would live with it were the four-person support team, two customer success managers, and the engineers building proration in six weeks. The default plan was a thirty-minute all-hands where leadership explains the rationale and asks if there are questions — and where the only questions asked are logistical, because nobody asks the real question in a thirty-minute all-hands.

We ran a Conversation Cafe instead, the week before the customer emails went out.

The structure

Conversation Cafe is built for exactly this: a group absorbing something disruptive they didn’t choose and can’t reverse. Tables of five to seven, a talking object per table, four rounds — one uninterrupted circuit on the invitation question, a second circuit responding to what they’ve heard, then open conversation with the object available in the middle, then a closing circuit of takeaways.

The agreements get read out first and matter more than usual when the topic is a decision people dislike: suspend judgment, listen to understand, speak from your own experience rather than on behalf of an absent group, use the object.

Ninety minutes, eleven people, two tables. I stated the agreements and then said essentially nothing for an hour and a half, which is harder than it sounds and is the job.

Running it on the pricing change

The invitation question: “When this lands with customers, what will be hardest — and what do you know that the rest of us don’t?” The second clause was deliberate. Support sits on an enormous amount of specific knowledge that never reaches a pricing decision, because nobody has built a route for it. Asking for it explicitly, in a format with uninterrupted airtime, is a route.

Round one was braced-for-impact — people naming accounts they expected to lose. Round two produced the finding. One CSM, holding the object, worked out loud through something she’d half-known for months: a large share of our biggest accounts have seat counts in the HR system that are wildly wrong, because nobody ever had a reason to clean them up. Offboarded employees stay flagged active for months. Seasonal staff at two retail customers triple the headcount every November.

Under flat tiers that was a data hygiene curiosity. Under per-employee pricing it’s an invoice saying a customer owes us for four hundred people who left last year — and the first thing that customer does is not pay it, and the second is question every number we’ve ever shown them.

Nobody in the pricing decision had known that. It surfaced because a talking object gave one person ninety uninterrupted seconds to finish a thought she’d never had reason to finish.

The open round surfaced the other half: support’s real fear was less about churn than about explaining a change they’d had no part in, to customers who’d ask whether it was fair. One newer rep said she didn’t know what she was allowed to say if a customer asked whether she thought the pricing was reasonable. That got a real answer that afternoon and became a line in the internal FAQ.

We shipped a seat-count reconciliation tool and a ninety-day grace period on inflated counts before the change went live. Both came out of that room. The decision itself didn’t change, and I said so at the top — but the rollout did, substantially.

Why the format survives a non-negotiable topic

The obvious objection is that running a listening structure on a settled decision is manipulation with extra steps. It is, if you pretend the decision is open. It isn’t, if you’re explicit about what’s fixed and what isn’t.

What Conversation Cafe does here is separate two things a normal all-hands fuses: processing the news and reacting to the plan. In a Q&A the only available move is to challenge the decision, so people either challenge it uselessly or say nothing. In a talking-object circle the frame is sense-making, so what comes out is knowledge rather than objection — and knowledge is what you can act on.

Round one drains the immediate reaction, which people need to say before they can say anything else. Round two, after hearing everyone, is where operational knowledge appears. Skip round two and you get vented feelings and no seat-count discovery.

When to skip it

Skip it if you’re not prepared to change anything. If rollout, messaging, timeline, and tooling are all locked, this is a ninety-minute performance of consultation and everyone will identify it as one by round three. Skip it too if the group can’t split into real tables of five to seven, unless you have a facilitator per table whose job is the agreements and nothing else.

And don’t run it instead of explaining the decision. The rationale has to be delivered plainly first, by someone who owns it. Conversation Cafe metabolizes news; it isn’t a way of avoiding delivering it.