Governance, in an operating-model sense, is the standing system of who decides what and who else is involved, made explicit, rather than renegotiated fresh in an email thread every time a decision recurs. The single most useful tool for this is a RACI matrix, assigning exactly one of four roles to every person or function touching a decision: Responsible (does the work, drafts the analysis), Accountable (owns the decision, the one name on the line if it goes wrong, and the only role permitted to say yes), Consulted (gives input beforehand, typically holding relevant expertise), and Informed (told after, with no input role beforehand).
The rule that makes a RACI work, and the one beginners violate almost every time: every decision gets exactly one Accountable. Not zero, an unowned decision stalls, everyone waiting for someone else, a failure Practice P12 drills directly. And not two or more, multiple Accountables is not more careful governance, it is less, because each extra "A" is another full sequential veto gate, and four sequential vetoes take roughly four times as long as one real owner, with no gain in accuracy, since none of the extra approvers is actually more informed. Worked Example 1 shows this pattern costing real weeks; Worked Example 3 rebuilds the decision with a single Accountable and shows what that change buys back.
Advertisement
What does a delegation-of-authority table add?
A RACI only tells you who; a delegation-of-authority (DOA) table tells you how much, tying the "who" to a threshold (typically a rupee or dollar exposure) below which one role decides alone, above which it escalates to a more senior Accountable. A well-built DOA table is tiered (multiple thresholds, not one blanket rule), specific (a stated number, not "use judgment"), and revisited on a cadence (a threshold sized for 45 stores silently mis-delegates once the same chain runs 180, quietly reintroducing the delay a redesign was meant to remove). Worked Example 3 builds a three-tier DOA table and computes the blended cycle-time improvement it buys.
| Role | Old ("broken") pattern | Fixed pattern |
|---|---|---|
| Person who drafts the analysis | R, with no real authority attached | R, with delegated authority up to a stated threshold |
| First-line manager | A (one of several sequential approvers) | Sole A within their DOA tier |
| Specialist functions (Finance, HQ policy owners) | A or a blocking gate on every single case | C, sets the guardrail policy on a standing cadence, consulted on exceptions only |
| Senior executive | A on everything, regardless of stakes | A only above the top DOA threshold; I (a dashboard) below it |
How does governance keep decisions moving?
Governance is the RACI-and-DOA artifact, and it is also the cadence of standing forums that keep decisions moving without reopening the whole matrix each time, a regular operating review where routine decisions live, a periodic policy-setting session where Consulted functions update their guardrails, and an escalation path used rarely precisely because the DOA table above it is doing its job. Practice P11 designs exactly this cadence for a fresh scenario.
{
"skill": "Drafting a first-pass RACI matrix and delegation-of-authority table for a real, recurring decision",
"use": "Have an AI tool generate a structured RACI/DOA skeleton — roles down the side, the decision's sub-steps across the top, a first-pass threshold ladder — that you then edit for a single Accountable per row, real threshold numbers, and a governance cadence, never a matrix that ships as the AI first drafted it.",
"tools": ["Chat assistants — Claude, ChatGPT, Gemini, for the matrix skeleton", "Spreadsheet tools — for the actual DOA threshold table with real numbers"],
"prompt": "For the decision '<DECISION, e.g. approve a regional promotional pricing plan>' at <ORGANIZATION>, draft a RACI matrix (Responsible/Accountable/Consulted/Informed) across the roles <LIST ROLES>, and a three-tier delegation-of-authority table by <RUPEE/DOLLAR> exposure. Flag any row where more than one role looks like it should be Accountable, and ask me to resolve it rather than picking one yourself.",
"verify": "Every Accountable assignment and every DOA threshold is a real business judgment about who bears the consequence of a wrong call — verify each against how the organization actually works (not a generic template) and confirm exactly one Accountable per decision before this leaves your desk; recompute any cycle-time claim yourself (Guardrail 1-5).",
"diy": "You must already know the single-Accountable rule and be able to spot a 'too many A's' pattern by eye, unaided, before you can tell whether an AI-drafted matrix actually fixed a broken decision or just relabeled the same broken one.",
"market": "IN"
}