1-2-4-All: Untangling an Approval Workflow Nobody Agreed On

The third time I ran 1-2-4-All as a workshop opener, it was for a leave-balance product’s approval-workflow revamp, and it taught me something the first two runs hadn’t: the structure is just as useful when the room’s problem isn’t one dominant voice, but three different stakeholder groups who each have a completely different mental model of what “the approval workflow” even is, and no shared vocabulary to notice they’re talking past each other.

The structure, quickly

For readers new to this series: 1-2-4-All is a four-step cascade — one minute of silent individual writing on a single question, two minutes in pairs merging answers, four minutes as pairs combine into fours and converge again, then a whole-group share where each foursome contributes only its strongest idea. Twelve to fifteen minutes total regardless of room size. The first post in this series has the full mechanics and a worked example; here I want to focus on what happens when the room’s problem is fragmentation of perspective rather than domination by one voice.

Three rooms in one room

An approval-workflow revamp for leave requests sounds like a single, coherent feature, but it’s really three different products wearing one UI, and the kickoff meeting had one representative from each: a people-manager who approves requests, an individual contributor who submits them, and an HR operations lead who owns policy and compliance.

Ask this room an open question — “what should the new approval workflow do?” — and you get three answers that sound like they’re describing different systems, because functionally they are. The manager talks about needing to see team coverage at a glance before approving, so they’re not accidentally approving five people off the same week. The employee talks about wanting to know why a request is pending for four days when it should be a same-day yes-or-no. The HR lead talks about needing every exception and override logged for an audit that happens twice a year and is currently reconstructed by hand from email threads.

In an unstructured discussion, whichever of these three speaks first and most fluently tends to set the frame the other two react to, and the meeting often ends up scoped around that first framing with the other two perspectives folded in as afterthought line items — “and also maybe some kind of audit log” — rather than as first-class requirements with their own weight.

Running it to force parallel, not sequential, surfacing

I used the prompt: “Think about the last leave request you were personally involved in, in whatever role. What was the one moment where the current process made you least confident it was doing the right thing?” The word “confident” was deliberate — it invites a judgment about trust in the process rather than a complaint about a specific UI element, which pulls each person toward describing their actual mental model of what the workflow is for, not just what’s annoying about the screen in front of them.

The silent minute produced three genuinely different but individually coherent answers, exactly as expected — but because the writing happened simultaneously and privately, none of the three was shaped by having heard the others first. That’s the part sequential discussion can’t replicate: in a normal meeting, the second and third speaker’s answers are already, whether they intend it or not, positioned relative to the first. Here, all three were pure.

The pairing round is where something useful happened that wouldn’t happen in a three-person unstructured conversation: with six people in the room (two per stakeholder group), the pairing was mixed — a manager paired with an employee, not manager-with-manager — so the two-minute synthesis forced an actual translation exercise between two different mental models rather than two similar ones reinforcing each other. The manager-employee pair converged on a shared idea neither had written alone: that “pending for four days” and “no visibility into team coverage” were the same underlying gap — a lack of any intermediate status the requester could see, like “your manager is checking coverage before deciding,” which would make the wait legible instead of silent.

By the whole-group share, the room had three distinct, well-formed needs on the board — visible team-coverage context for the approver, legible in-process status for the requester, and structured exception logging for HR — none of which had crowded out the others, because the structure guaranteed each stakeholder’s silent-minute answer survived intact into at least the pairs round before any synthesis or trade-off conversation began.

Why sequencing matters more than usual here

In the pipeline-feature kickoff, 1-2-4-All’s job was preventing one voice from dominating. Here its job was different: preventing the first framing from becoming the only framing, in a room where all three perspectives were individually valid and none was wrong — they were just describing three different failure surfaces of the same workflow. A structure that surfaces ideas sequentially, even a well-run round-robin, still has an ordering effect: whoever’s model gets described first becomes the lens the second and third contributions get evaluated against. Simultaneous silent writing removes ordering entirely. That’s the specific value this structure added to a multi-stakeholder kickoff that a skilled facilitator running an unstructured discussion, even one determined to be fair to all three groups, would have had a hard time replicating by force of will alone.

The workflow that got scoped afterward had three named pieces mapped directly to the three surfaced needs, rather than one piece with two footnotes — which mattered later, because it meant none of the three stakeholder groups could reasonably say their need had been deprioritized without a conversation; it was visibly present in the plan from the start.

Failure modes and when to skip it

This structure struggles when the stakeholder groups in the room are wildly unequal in size or seniority — if it’s one HR lead against five line managers, the pairing and fours rounds will statistically dilute the minority perspective even though the silent minute protected it, simply because there are more manager-shaped ideas to average against. In a lopsided room like that, consider running separate 1-2-4-All sessions per stakeholder group first, then a single facilitator-led synthesis across the three outputs, rather than mixing bodies in the room from the start.

It also isn’t the tool for resolving an actual conflict between stakeholder needs once they’re identified — “the manager wants coverage visibility, but HR’s policy says coverage data can’t be shown to avoid bias accusations” is a real trade-off that needs deliberate discussion, not another round of silent writing. 1-2-4-All surfaces the needs; it doesn’t adjudicate between them when they collide.