Hoshin Kanri for executive committees: make decisions stick

Most executive committees do not lack decisions — they lack follow-through. The minutes live in a document nobody reopens, actions are assigned verbally, and the next meeting re-discusses the same topic with the same slides. Three symptoms tell you the committee needs a deployment frame rather than another meeting:

• Decisions get re-litigated because nothing records what was decided, by whom, and against which objective.
• Each department presents a plan, and nobody can say which company objective a given plan serves — or whether two plans quietly contradict each other.
• The KPI review is a tour of a dashboard, without ever asking which strategic bet each number is supposed to prove.

The X-Matrix puts the entire strategy on one page: 3-5 year breakthrough objectives (North), annual objectives (East), improvement initiatives (South) and KPIs (West), connected by correlation marks in the four corners. The point is not the diagram — it is the negotiation it forces. Each element has an owner around the table, and each correlation mark is a claim someone has to defend: "this initiative actually serves that objective".

That negotiation — the catchball — is what makes the matrix binding where a slide deck is only descriptive. Disagreements surface before the year starts, in front of the whole committee, instead of resurfacing in March as passive resistance.

Hoshin Kanri structures the meetings around the matrix in a five-level hierarchy — from the annual strategic review down to weekly syncs. For an executive committee the working set is usually three: the annual review that rebuilds the matrix, a monthly steering meeting, and a weekly sync on off-track items.

Every agenda is generated from the matrix itself: KPIs off trajectory, blocked initiatives, decisions pending since the last meeting. Decisions and action items are recorded against matrix elements — so "what did we decide about X?" always has an answer, with a date and an owner.

Connected to the matrix through MCP (two minutes with OAuth from claude.ai or Claude Desktop), an AI agent does the work a chief of staff used to do around the committee:

• Before the meeting: runs the weekly sync — pulls progress from Linear or Jira, updates initiative statuses, drafts the agenda with off-track KPIs and pending decisions.
• During the meeting: captures ideas and decisions into the matrix inbox mid-conversation, so nothing depends on someone remembering to write minutes.
• Between meetings: any committee member asks their own assistant "where are we on the retention objective?" and gets the live answer — instead of asking a team to build a deck.
• Continuously: flags drift — work in the execution tools that maps to no initiative, initiatives that no longer serve any objective.

Week 1: a half-day workshop builds the first matrix from the free template — 3 to 5 long-term objectives, no more. Weeks 2-3: catchball with the n-1 level; annual objectives and initiatives get owners. Before locking: the completeness score audits the matrix — orphan initiatives, unmeasured objectives, missing vision-to-KPI chains. Then the cadence starts.

By the second monthly review, the agenda writes itself and the meeting is arbitration instead of status collection. That is the test of whether the method has taken: the committee spends its time deciding, not reconstructing what happened.