What an AI Operations Coordinator Actually Looks Like

"We need an ops person" is often really "we need the coordination work done." Requests routed, tasks created when deals move, nobody forgetting follow-ups, sales calls walked into prepared. That coordination layer is exactly what automation is good at — and it doesn't need to sleep.

The problem

A global climate advisory firm had requests arriving through chat threads and inboxes, deal-stage changes that should have spawned task playbooks but relied on memory, overdue tasks that nobody chased, and sales calls that started with zero background on the person across the table.

The build

Seven connected n8n workflows, each owning one job. Requests flow in through a form and become categorized tasks in Notion in under three seconds. When a deal changes stage in HubSpot, the matching task playbook spawns automatically. A nightly scan flags overdue and stale tasks with direct comments to the owner.

The two AI pieces earn their place: when a meeting gets booked, a 12-step research agent assembles a formatted brief on the company and attendee before every sales call. And a daily lost-deal agent scans the CRM, judges which closed-lost deals are worth re-engaging, and emails the owner a simple Approve/Skip decision. Humans stay in the loop exactly where judgment matters.

The gotcha

The temptation with AI agents is to let them act. The lost-deal agent doesn't send a single email to a prospect — it recommends, a human approves. That one design choice is the difference between an ops tool a COO signs off on and a liability nobody trusts. Autonomy is a dial, not a switch.

The result

The system was built and presented live to the firm's Co-Founder & COO in under five hours of build time. Intake, task playbooks, overdue nudges, pre-call research briefs, and lost-deal recovery — the coordination layer of an ops role, running on a schedule.