COO — Charter

If the CEO decides what, the COO makes it happen. In an AI-operated company that job shifts from coordinating people to designing and maintaining the system that coordinates the work — because much of the coordination itself is now automated, and a badly designed workflow now fails faster and at larger scale.

What this role owns#

Execution. Whether what was decided actually gets done, at the agreed quality, by the agreed date.

The operating cadence. Which meetings exist, what each is for, and — more usefully — which do not exist any more because their output is now generated.

Cross-function throughput. Work rarely stalls inside a department. It stalls between them, and nobody owns the gaps by default.

Resource allocation across departments: people, agent capacity, and budget for both.

Process design. Including the honest judgement of which processes should be automated, which should be simplified first, and which should be deleted.

KPIs#

Handoff wait time is the one most organisations never measure and most would benefit from measuring. Work is usually waiting, not being worked on.

AI agents in this function#

Workflow monitor — watches work moving between departments and flags where it stops. Turns "we're behind" into "eleven days were lost waiting for legal review".

Capacity forecaster — projects load against available capacity and flags the constraint before it bites.

Process mining agent — reconstructs how work actually flows from the systems of record, which is reliably different from how the documented process says it flows.

Meeting reduction agent — produces the status a meeting existed to share, so the meeting can stop.

What stays human: deciding what to stop doing, resolving a conflict between departments, changing someone's role, and judging whether a process failure was a system problem or a person problem. That last distinction is where most operational damage is done by getting it wrong.

SOPs#

• Weekly execution review — exceptions and blockers only. Anything green takes no time.
• Escalation protocol — what escalates, to whom, and the response time promised.
• Handoff standard — every handoff states what is being passed, what "ready" means, and who accepts it. Most handoff waste is an undefined acceptance criterion.
• Capacity planning — monthly, covering agent spend as well as headcount.
• Process change — measure the current state before changing it, or you cannot tell whether the change helped.

Templates#

Project Plan · SOP · Incident Report · capacity plan · handoff definition.

Workflows#

In: decisions from the CEO · department status · escalations · capacity signals · customer impact reports.

Out: allocated resources · resolved blockers · process changes · exception reports upward.

Handoffs: every department, in both directions. This role is the hub, which is exactly why the handoff standard matters more here than anywhere else.

FAQ#

What is the difference between the COO and the PMO?#

The PMO governs how projects are run and reports on them. The COO owns whether the organisation delivers, including everything that is not a project — support, operations, recurring work. In a small company one person may do both; the accountabilities are still different.

What is the first thing to automate?#

The recurring work that a person does the same way every time and that blocks someone else while they do it. Status collection is nearly always the first, because it is high-frequency, mechanical, and its absence blocks decisions.

How do you know a process should be deleted rather than automated?#

Ask which decision it informs. If the answer is "none" or "it's for the record", automating it makes a pointless activity faster and more consistent. Delete it instead.

What breaks most often in an AI-operated company?#

Handoffs between an agent and a person. The agent completes its part and nothing tells anyone; work sits in a state nobody is watching. Every automated step needs a defined destination and an owner.

What else is coming for COO