ServiceNow RACI Matrix: Two People Accountable Is The Same As Nobody

Most platform programmes that stall are not short of effort, they are short of a decision, and the decision is short of an owner. The build is funded, the teams are staffed, the governance paperwork looks complete, and one call has been open for a month because nobody in the building is certain it is theirs to make.

I have watched this from inside enough programme reviews to treat it as a rule rather than a run of bad luck. When a ServiceNow programme misses a quarter, the post-mortem rarely finds a task that nobody did, it finds a question that nobody answered.

Here is the part that should bother you. The programmes that stall this way are usually the well documented ones. There is a roles matrix, workshopped and signed off, sitting in the project wiki. And the health check six months later still records the same finding, unclear roles and responsibilities. Roles cannot be documented and unclear at the same time, unless the document is answering a different question from the one the programme keeps asking.

Tasks get done, decisions get made, and the grid holds only one of them

A responsibility matrix allocates work, and it says nothing about who is allowed to end an argument.

Look at what the rows in a typical matrix actually are. Build the integration, test the workflow, sign off the release. These are pieces of work, and mapping them to people is genuinely useful, which is why the matrix has survived every attempt to retire it. Work that is visible, finishable and assigned tends to get done.

Now list what your own programme is actually stuck on, and you will find a different species of thing entirely. Whether the regional service desk keeps its own intake process or adopts the shared one. Whose record wins when two systems disagree about the same customer. Whether one department's exception to the standard way of working is granted or refused. None of these is a task, and nobody can start one, progress one or complete one. They are arguments, and an argument does not need a resource assigned to it, it needs a referee.

This is why bolting a steering committee on top of the matrix so often fails to help. A committee is where decisions go to be escalated, and a good one resolves what the delivery team genuinely cannot. But when the grid beneath it names no decision owners at all, every disagreement flows down the escalation path, the agenda bogs things down, and a body that meets monthly becomes the bottleneck for calls that should have taken a day.

One name is a decision, two names is a deferral

The moment a second name enters the accountable column, each owner acquires a reason to wait.

The delivery method most platform teams borrow their matrix from is blunt about this, its own setup guidance tells you to make sure no task carries more than one accountable owner. One, exactly, every time, and it allows no exceptions to that rule anywhere in its library.

Teams break the rule constantly, and they break it for motives that feel virtuous, because two names looks fairer, reads as collaboration, and on paper spreads the risk. In practice it concentrates the risk, because accountability is the willingness to be the person the question comes back to, and that willingness does not survive company. When two people share the accountable slot, each can see the other, each reasonably assumes the other is closer to the detail, and the call settles into the space between them. Neither of them is negligent, both of them are simply waiting, and the matrix records the arrangement as clarity.

Writing a committee's name in the slot is the same mistake in formal dress. A board that convenes monthly cannot own a call that recurs weekly, so the decision waits for the meeting, the meeting defers it for more analysis, and the matrix continues to insist that somebody is accountable. Committees exist to catch what named owners cannot settle, and they only work when those owners exist.

Consulted is a comfortable place to hide

Almost everyone in a stalled programme sincerely holds a role in the decision, and almost nobody owns it.

Run the test in the room, and you will see the shape of it. Ask who is involved in the stuck decision and every hand goes up. Ask whose decision it is and the hands come down, replaced by a silence in which several people look at one person, who is at that moment looking at someone else.

Change-management research has been circling a version of this for decades. A leader who supports an initiative is not the same as the leader who sponsors it, and organisations confuse the two constantly, because support is cheap and sponsorship has a price. The consulted column fills itself for the same reason, because being consulted carries influence without exposure, you shape the call, you sit in the room, and when it goes wrong the failure is not yours. Ownership offers the opposite deal, and so the accountable column stays thin unless someone is deliberately placed there and agrees, out loud, to stay.

This is why filling in the matrix at a workshop feels productive and changes so little, because the workshop only ever fills the columns people are happy to occupy.

Put the name where the consequence lands

The right owner has the information to judge the call and lives with its result. That is frequently not the most senior person in the room.

When teams do write a single name, they tend to copy it from the org chart. The most senior title takes the accountable slot by default, on the logic that authority rises with altitude. Good platform practice pushes the other way, the people administering a process day to day should not be the ones accountable for its strategy, and the owner of an operational call should not sit three levels above the information needed to make it.

A decision about how a shared intake process works belongs with the person who will answer for the queue it produces. A decision about whose data is treated as the truth belongs with the person who carries the consequence when the wrong record wins. Give those calls to a senior sponsor instead and you have not added authority, you have added distance. The stall reproduces itself at higher altitude, with the added cost that nobody below feels entitled to chase the owner for an answer.

Exhibit. Same people in the room, and the accountable slot decides whether the call gets made.

The consulted group is identical in both states, but in the first, the accountable slot holds two names or none, and the decision has been open for three weeks. In the second, it holds one name, and the decision is made, and nothing else about the room has changed.

Where this doesn't apply

Not every choice deserves this ceremony, and the cheap, reversible calls a team can undo next sprint should simply be made and moved past, and naming an owner for each of them is bureaucracy wearing the costume of rigour. At the other extreme, a genuinely strategic commitment the organisation cannot walk back may belong to a board acting as a single body, because there the point is collective consent rather than speed. The one-name rule earns its keep in the middle, on the recurring operational decisions a programme trips over week after week.

The bottom line

Do not rerun the workshop, and do not buy a better template. Take the decisions your programme has actually been stuck on this quarter and try to write one name against each of them. Where the name arrives in a second, you were never really stuck. Where it will not come, you have found the real work, and it is not documentation. It is a negotiation the organisation has been avoiding, and no matrix will hold that negotiation until somebody has it.

Two names in the accountable column is not shared ownership. It is a decision the organisation has quietly agreed not to make.

P.S. Watch what happens to decisions that get parked in your next programme review. A parked decision with an owner comes back with an answer. A parked decision without one just comes back.