Harvard Business Review's case for a chief of staff starts from a simple observation: CEOs need more support than an executive assistant can provide. Most of that support is rhythm. Someone checks every morning what moved and what stalled, routes new work to the right owner, chases what the team is waiting on, runs the meeting and writes the wrap. That rhythm is repetitive, it runs on data already in your tools, and it is easy to measure. That makes it the best first job for an AI employee, and the AI COO or AI chief of staff the most useful first hire.
1. Rhythm Is a Schedule, Not a Prompt
An AI COO that waits to be asked is an assistant. The job is defined by when it happens: a weekday-morning operating check, a routing pass as work arrives, an end-of-day wrap, a weekly look ahead. In Kavanah each of those is a schedule bound to the AI COO, running with its scopes and its model, and delivering its result by email to the people who need it. Anthropic's advice for agents applies directly: find the simplest design that works and add complexity only when it is needed. Start with one schedule and add the next when the first is read every day.
2. The Morning Check
The useful morning brief is short. What moved yesterday. What is overdue or stuck, and who owns it. What the team is waiting on from outside, and for how long. And the decisions only a person can make, listed rather than made. Kavanah's AI COO reads open tasks, overdue work and waiting-on items with only the tools in its scope, and the roster's This week view shows its own runs, skips and failures so you can see whether it is doing the job.
3. Routing and Chasing
Routing new tasks to owners against the team's limits is a good AI job, because it is rule-driven and the rule is yours. Working tasks is a bigger step: in Kavanah a task assigned to an AI employee is worked by it only once an admin turns that on, and the run borrows the access of the person who assigned it, never more, capped at 20 runs an hour per workspace. Gartner expects at least 15% of day-to-day work decisions to be made autonomously by agentic AI by 2028; routing is a sensible place for some of yours to start.
4. The Meeting and the Wrap
An AI COO can chair a daily executive meeting of AI officers, once exec meetings are enabled: each officer gives an update from read-only tools, they discuss without tools, and the chair closes with action items that become tasks with owners. At the end of the day the wrap reports what closed, what slipped and how the day moved the number the team cares most about. Neither replaces the human meeting; both make it shorter.
5. What Stays With People
Priorities, trade-offs and anything a person is answerable for. An AI COO in Kavanah never makes, approves or records a decision reserved for a person. When it needs one, it files a single approval request on the task, and with the daily decision window on, the reviewer sees all open asks in one digest a day, with one escalation after 48 hours. No agent tool can approve or reject an ask.
Give an AI COO the rhythm and keep the judgement. Start with a morning check, read it for a week, then add routing, the meeting and the wrap. For how AI employees work in Kavanah and how other vendors compare, see our AI employees page.
