The RACI Matrix

The decision document captured what was decided and why. But a decision that lands on a team and then dissolves into “wait, who’s actually doing this?” hasn’t really been made — it’s just been announced. This page hands you the tool for the next question: once there’s work to do, who does what. It’s a grid that maps every task against every person and marks, for each pairing, exactly one of four roles. Fill it in and the two most common ownership failures — a task nobody owns, and a task three people are quietly doing at once — become impossible to miss.

The RACI matrix isn’t a heavyweight process. At its smallest it’s a table you sketch in five minutes at the start of a project. Its value isn’t the grid itself; it’s the conversation the grid forces — the moment where “I assumed you had it” and “I thought that was your call” get said out loud before they turn into a missed deadline.

What it’s for

RACI stands for four roles. The whole tool is just the discipline of assigning them, one row (task) and one column (person) at a time.

• R — Responsible. The person (or people) who actually do the work. They write the copy, run the tests, make the calls, stack the boxes. There can be more than one R on a task.
• A — Accountable. The single person who answers for the outcome — the one who says “yes, this is done and done right,” and the one on the hook if it isn’t. Every task has exactly one A. Not zero, not two. This is the rule that makes RACI work.
• C — Consulted. People whose input you actively seek before acting — two-way. You ask them, they answer, their view shapes the work. A specialist, a stakeholder, someone whose domain you’re touching.
• I — Informed. People you tell after the fact — one-way. They don’t shape the decision; they just need to know it happened so they’re not surprised.

The single most important rule is worth repeating on its own: each task has exactly one Accountable person. Responsible can be shared. Consulted and Informed can be crowds. But accountability cannot be split — the moment two people are accountable for the same outcome, each can reasonably assume the other has it, and it falls between them. One throat to choke, as the old phrase goes; more usefully, one name to thank when it goes right.

R  Responsible   does the work            (can be more than one)
A  Accountable   answers for the outcome  (EXACTLY one per task)
C  Consulted     asked before — two-way   (can be several)
I  Informed      told after — one-way     (can be many)

The template

Tasks go down the left. People (or roles) go across the top. In each cell, put the letter — or leave it blank if that person has no part in that task. A blank cell is a real answer: it means “this person doesn’t touch this.”

                    │ Person A │ Person B │ Person C │ Person D │
────────────────────┼──────────┼──────────┼──────────┼──────────┤
 Task 1             │    A     │    R     │    C     │    I     │
 Task 2             │    R     │    A     │          │    C     │
 Task 3             │    I     │    C     │    A     │    R     │
 Task 4             │    A     │    R     │    R     │    I     │
────────────────────┴──────────┴──────────┴──────────┴──────────┘

Rule check for every row:  exactly one A?  at least one R?

In plain markdown, the same grid is just a table — which is the form you’ll usually paste into a doc:

| Task              | Person A | Person B | Person C | Person D |
|-------------------|----------|----------|----------|----------|
| Task 1            | A        | R        | C        | I        |
| Task 2            | R        | A        |          | C        |

How to use it

1. List the tasks first, before the names. Break the work into the real chunks that need an owner — not “do the project,” but “book the venue,” “sign off the budget,” “write the announcement.” If a task is too big to point at one A, split it.
2. Add the people across the top. Use names, not just titles — RACI is about who actually, not who nominally. Include everyone who touches the work, even lightly.
3. Fill in one row at a time, A first. For each task, ask “who answers for this?” and put a single A. Then “who does it?” (R). Then “whose input do we need first?” (C) and “who just needs to know?” (I).
4. Run the two checks on every row. Exactly one A? At least one R? A row with no A is an orphaned task. A row with two A’s is a fight waiting to happen.
5. Read the columns too. A person with A’s on everything is a bottleneck. A person who is C on nearly every row is a drag on speed (see the diagnostics below).
6. Share it and let people argue. The grid is a draft until the people in it have seen their own row and column and agreed. Disagreement here is cheap; disagreement mid-project is expensive.

A filled-in example

Here is a product-launch project for a small team. Notice that each row has exactly one A, that R’s are sometimes shared, and that blank cells are used freely — most people don’t touch most tasks.

| Launch task                | Priya (PM) | Sam (Eng) | Dana (Mktg) | Leo (Sales) | Ravi (VP) |
|----------------------------|------------|-----------|-------------|-------------|-----------|
| Finalize launch date       | A          | C         | C           | I           | I         |
| Ship the feature           | I          | A / R     | I           |             | I         |
| Write the announcement post| C          | I         | A / R       | I           | I         |
| Approve pricing            | C          |           | C           | C           | A         |
| Brief the sales team       | I          | I         | R           | A           | I         |
| Update the help docs       | A          | C         | R           |             |           |
| Go / no-go call            | A          | C         | C           | C           | I         |

Read a couple of rows in English to feel how it works:

• Finalize launch date: Priya answers for it (A). She consults Sam on whether the build will be ready and Dana on marketing timing (C). Leo and Ravi are just told the date (I).
• Approve pricing: This is Ravi’s call — he’s Accountable (A) — but he consults Priya, Dana, and Leo first (C). Nobody is marked R because approving is the work, and the A carries it.
• Ship the feature: Sam is both Accountable and Responsible (A / R) — he does it and answers for it. Everyone else is simply Informed.

The same grid works for a hiring process — swap the columns for the interview panel:

| Hiring task                | Nadia (Hiring Mgr) | Tom (Recruiter) | Panel | Ravi (VP) |
|----------------------------|--------------------|-----------------|-------|-----------|
| Write the job description  | A / R              | C               |       | I         |
| Source candidates          | C                  | A / R           |       | I         |
| Screen resumes             | A                  | R               |       |           |
| Run interviews             | C                  | I               | A / R | I         |
| Make the final decision    | A                  | I               | C     | C         |
| Send the offer             | I                  | A / R           |       | C         |

Notice the panel is Consulted on the final decision, not Accountable — many voices, one owner. That single A on “Make the final decision” (Nadia) is what stops a hire from stalling in a committee that can never quite agree.

Reading the grid as a diagnostic

The real power of RACI shows up after you fill it in. The pattern of letters is a map of how your work will actually flow — and its problems are visible at a glance.

• A row with no A — an orphaned task. Nobody owns the outcome. This is the task that gets discovered three days late, when everyone assumed someone else had it. Every blank-A row is a landmine; find them before the project does.
• A row with two A’s — a collision. Two people each think they have the final say. Either they’ll clash, or — worse — each will assume the other is handling it and neither will. Split accountability is the same as no accountability. Force it to one.
• A column full of C’s — a decision bottleneck. If one person is Consulted on almost every task, the work can’t move without waiting on them. That’s a person who has become a checkpoint for everything and a completer of nothing.
• Too many C’s across the whole grid — slow decisions. When lots of tasks carry lots of Consulteds, every action needs a round of asking-first. That feels inclusive, but it’s how a two-week job becomes a two-month one. Ask of each C: does their input actually change what we’d do? If not, demote them to I.
• A person with only I’s — are they even on this project? Sometimes that’s correct (a senior sponsor who just wants updates). Sometimes it’s a sign you’ve included someone out of politeness who doesn’t need to be there at all.

The common trap — RACI-ing the small stuff

The most common failure isn’t filling the grid in wrong — it’s reaching for the grid at all when a thirty-second conversation would do. A two-person task (“Can you handle the slides while I book the room?”) does not need a matrix; drawing one is slower than the work. RACI earns its overhead when ownership is genuinely muddy — several people, crossing teams, unclear who decides — and it becomes pure ceremony when applied to work small enough to sort out by talking. As the toolkit overview puts it: overhead should match the stakes. If you can resolve who-does-what in one clear sentence, say the sentence and skip the grid.

In practice — the launch that shipped twice

A small e-commerce team once ran a promotion where both the marketing lead and the store owner believed they were sending the customer email. Neither had marked it down; each assumed the other had it. The email went out twice, four minutes apart, and the second one carried the wrong discount code. The fix afterward wasn’t a new tool or a scolding — it was a five-minute RACI on the next campaign, where the single A on “send the customer email” made the double-send structurally impossible. The grid didn’t add process; it removed an argument that hadn’t happened yet.

The evidence — accountability and the bystander effect

Social psychology has a well-studied finding often called the bystander effect or diffusion of responsibility: when several people could act, each is less likely to, because responsibility feels spread thin — everyone assumes someone else will step in. Studies generally find that naming a specific person (“you, in the blue shirt, call for help”) dramatically raises the odds that action is taken. RACI’s one-A rule is that finding applied to work: a task with a named owner gets done; a task everyone “sort of” owns gets orphaned. The letter A is just a way of pointing at the person in the blue shirt.

Case study — NASA and the ambiguity of ownership

Investigations into major organizational failures repeatedly surface the same root cause: not that a warning was missing, but that no single person clearly owned the decision to act on it. The 1986 Challenger review (the Rogers Commission) and later the 2003 Columbia board both described communication and responsibility gaps where concerns were raised but ownership of the go/no-go call was diffuse — engineers flagged risks, managers assumed the risk had been someone else’s to weigh. The lesson isn’t that a RACI grid would have prevented a tragedy; it’s that in exactly the moments that matter most, who is Accountable for this call must be unambiguous and known in advance — never sorted out in the heat of the moment.

Try this

Take one real piece of upcoming work where ownership feels even slightly fuzzy — a launch, an event, an onboarding, a migration. Spend fifteen minutes filling in the grid: tasks down the side, names across the top, one letter per cell. Then run the checks: does every row have exactly one A? Are there columns drowning in C’s? Share the draft with the people in it and let them correct their own rows. If any single answer surprised someone, the grid has already paid for itself.

How it connects

RACI is the ownership layer under several other tools. It plugs directly into the project charter, which names the work the matrix then assigns; it clarifies who the single Accountable owner is behind a decision document; and it draws on the ideas in ways of working about making responsibility explicit. When the weekly status report asks “what do you need, and from whom,” a good RACI is often where that “whom” was decided.