Turning a workshop's sticky notes into a RACI that actually resolves arguments

Background

A RACI matrix built live in a workshop is a social document before it's a management one. Someone senior says "I suppose I should be accountable for that," a couple of quieter people get marked Consulted because nobody wants to leave them off the page, and the facilitator moves on rather than push back on the third task in a row with two names under Accountable. The result looks finished and resolves nothing - the first time a decision actually needs making, two people who were both marked Accountable find out in the room, not before it.

The fix isn't a better workshop technique. It's building the RACI somewhere the model can be blunt in a way a facilitator in the room with those people usually won't be: after the fact, from the raw notes, explicitly asked to flag every row with more than one Accountable or with nobody accountable at all, rather than smoothing it into something that merely looks tidy.

Setup

The shape that gets a usable RACI out of messy inputs:

• Role - a RACI facilitator whose job is to surface conflicts, not hide them.
• Context - raw, unedited notes from the room: who said what, who volunteered for what, where two people both put their hand up.
• Inputs - the actual task list and whatever assignments came out of the discussion, contradictions included.
• Task - build the matrix, but explicitly interrogate anything that doesn't resolve to exactly one Accountable per task.
• Constraints - it isn't allowed to quietly pick a winner between two claimed owners. It has to flag it back to you.

Building the matrix from raw notes

You're building a RACI matrix from workshop notes, and your job is to
surface unresolved ownership, not paper over it.

Context: I'll paste raw notes from a workshop - task names, who spoke about
each one, and any ownership that was said out loud. Some tasks will have
been claimed by more than one person, or by nobody.

Inputs:
- Tasks discussed: [LIST, AS THEY CAME UP - DUPLICATES AND ALL]
- Raw notes per task: [WHO SAID WHAT, PASTE VERBATIM]
- Known team/roles in the room: [LIST]

Task: for each task, assign Responsible, Accountable, Consulted, Informed.
Accountable must be exactly one name. If the notes support more than one
candidate for Accountable, or none, list the task under a separate
"Unresolved" section instead of guessing - don't default to the most
senior person in the room just because they spoke last.

Constraints: every task from the input list appears exactly once. No task
silently dropped for being unclear - unclear tasks go in "Unresolved" too,
named as such.

Turning the unresolved list into decisions, not more discussion

The output above always produces an "Unresolved" section. Don't take it back to a workshop - it took a whole session to get to this ambiguity, and another session raising the same names in front of each other again solves nothing new. Take it to whoever the RACI ultimately reports to.

You're helping someone decide unresolved ownership from a RACI exercise -
without a second workshop.

Context: these tasks came out of a mapping exercise with either two people
claiming Accountable or nobody claiming it. I need a recommendation and the
reasoning behind it, not a repeat of the ambiguity.

Inputs:
- Unresolved task and the competing/absent claims: [TASK - CANDIDATE A vs
  CANDIDATE B, or TASK - NO CLAIMANT]
- What each candidate already owns elsewhere: [LIST, IF KNOWN]
- Anything about the task that makes one candidate the obvious fit
  (skillset, existing related ownership, capacity): [DESCRIBE]

Task: recommend one Accountable owner per task, with a one-line reason a
non-technical sponsor could act on without needing the full context
rebuilt for them. Where you genuinely can't recommend one from the inputs
given, say so rather than picking arbitrarily, and name exactly what extra
information would settle it.

Constraints: one recommendation per unresolved task, or an explicit "can't
recommend, here's why" - never a shrug with no next step attached.

Sanity-checking the finished matrix for overload

A RACI can be internally consistent - one Accountable per row - and still be unworkable, because it quietly makes one person Accountable for everything.

You're checking a finished RACI matrix for a single point of failure.

Context: I'll paste the finished matrix. Check the distribution of
Accountable and Responsible roles across people, not just that every row
resolves.

Inputs:
- Finished RACI matrix: [PASTE]
- Roughly how much capacity each person already has for this
  (full-time on it, part-time, this is on top of their day job):
  [LIST]

Task: flag anyone Accountable for more tasks than their stated capacity
plausibly supports, and anyone who appears in Responsible or Accountable
for a task clearly outside what their role would normally cover. Suggest,
for each flagged case, who else in the matrix could realistically absorb
some of it.

Constraints: don't just report a raw count of rows per person - weigh it
against the capacity you were given, since three small tasks and one huge
one aren't the same load even though they're both "three rows".