kenmulford RACI Matrix
Generate a Project+ RACI .xlsx. Draft-first. Only interview if the user explicitly asks.
Hard rules
- Never invent people or activities. Unknown cells →
[TBD]. Use exact user wording for activity names. - Reporting-line managers are NOT columns. "Marcus reports to Alicia" → Alicia gets no column. Only people the user described as doing project work. Sponsor is the exception. Check every column before running the script.
- RACI integrity: every row has exactly one A and ≥1 R. R and A may be same person (use
R/A). If the Accountable is unknown, set that cell to[TBD]— the script flags it on the Draft Gaps sheet. - "No team yet" → single column
[TBD — team unknown]with all cells TBD. Do not invent role columns from activity keywords (e.g. "finance dashboard" ≠ "Finance Lead"). - Don't expand "the team" into invented names. "The dev team deploys" → one column "Dev Team" with R.
Verb → letter
- builds / does / runs / executes / writes → R
- approves / signs off / has final say → A
- reviews / consulted / advises → C
- informed / notified / FYI → I
The doer is R, the approver is A — they are usually different people. "Ken drives scope, Mei approves scope" → Ken=R, Mei=A (one row, one A). Do not put A on both.
"I'm accountable overall" is PM framing, not a per-row A. Do not propagate an A into every row just because the PM said they're accountable for schedule/scope. PMs use "accountable" loosely to mean "ultimately on the hook" — a RACI A is narrower: the single person who signs off on this specific deliverable. If the user said "I'm accountable for schedule and scope," the PM gets at most two As: one row for "Schedule management" and one for "Scope management" (if those appear as activities). They do NOT get an A on "Build the data export tool," "Run the migration," or "Validate ticket history" — those have their own approvers.
Worked example (the trap): User says "Marcus owns the build. Priya approves cutover. I'm Ken, accountable for scope and schedule overall."
- ❌ WRONG: Build=Marcus R + Ken A, Migration=Marcus R + Ken A, Cutover=Marcus R + Ken A + Priya A (2 As), ...
- ✓ RIGHT: Build=Marcus R/A, Migration=Marcus R/A, Cutover=Marcus R + Priya A, Scope management=Ken R/A, Schedule management=Ken R/A.
Before running the script, scan every row: if the same person holds A across more than 2–3 rows and the user did not explicitly name them as approver on each, you are propagating. Stop and re-read the verbs.
Run it
python3 scripts/create_raci.py --output "<dir>/<ProjectName>_RACI_v1.xlsx" <<'JSON'
{ ...schema... }
JSON
Script validates RACI integrity, highlights TBDs, writes Draft Gaps + Integrity Issues sheets, and emits open_questions.md on many gaps.
JSON schema
{
"header": {
"project_name": "string",
"project_manager": "string or [TBD]",
"sponsor": "string or [TBD]",
"date": "YYYY-MM-DD",
"version": "v1"
},
"people": ["Column header 1", "Column header 2"],
"activities": [
{
"name": "activity / deliverable",
"phase": "optional phase string or null",
"assignments": {
"Column header 1": "R | A | C | I | R/A | [TBD] | \"\""
}
}
]
}
people = column headers in order. Every assignment key must match a people entry exactly. Empty string = not involved. Keep activity names ≤ 60 chars.