When our COO retired, she suggested we rebuild the role as agents.

The situation

In May 2026, LMNTL's COO planned her retirement. She'd built and run the role for years. With the date on the calendar, she was the one who proposed the experiment: take her playbook and rebuild it on GoodHelp, the same platform we ship to customers. If the agents could carry the work, fine. If they couldn't, we'd know exactly where the gaps were before her last day. The role spanned 53 distinct recurring responsibilities across finance, operations, and client delivery.

The plan

We took her up on it — mapped every responsibility and rebuilt it on GoodHelp, over a five-week transition planned against her retirement date:

• Finance, run by agents — daily cash, monthly close, billing, and a watch on vendor charges that flags anything unexpected before it clears.
• Growth, run by agents — partner outreach, new-client onboarding, and customer feedback routed straight to the roadmap.
• Operations, run by agents — the banking, billing, CRM, and help-desk systems the business already runs on, kept in sync, with problems surfaced before they become outages.
• Leadership visibility — a clear daily, weekly, and monthly digest of everything the agents got done, across every part of the business.

What we deliberately kept human

Not every responsibility should become an agent, and we didn't pretend otherwise. Relationship calls, configuration meetings, and the last-mile editorial sign-off stay with named people. Every high-risk financial action — vendor payments and the like — still needs a human approval before it goes through. That isn't a gap in the plan; it's the plan. The agents do the preparation and the routine work, and people keep the judgment calls.

The cutover

The transition is fully documented: a 53-row coverage audit where every responsibility has a named owner and a check on its status, plus an executive summary listing every risk and how we mitigated it. The cutover date — her last day, set by her — was held to the calendar.

The result

Coverage: 53 of 53 responsibilities, each with a named owner — most run as agents, a few became software, and the judgment calls stay with people. Evidence: a 53-row coverage audit and an executive summary documenting every risk and mitigation. Continuity: the same agents now run in the product our customers use.

We run our own back office on these agents, under the same audit trail we give our customers — the product we sell is the product we rely on.

These aren't demos. We run our own books on these agents every day, and LMNTL pays GoodHelp the same retail price any customer pays — no special internal treatment. Every agent we built for ourselves is one you can spin up too.