The back office worked. The business did not.

The business I shut down last week had one thing worth bragging about. The back office ran itself.

Three AI agents and a spreadsheet. That was the whole system.

I dropped prospect names into an Airtable from my phone. Ten seconds each. On a schedule, an agent researched each one: their website, their LinkedIn, their booking page, any public signal they had the problem I solved. It scored fit 1 to 5 against rules I wrote down, and drafted a one-line outreach angle referencing something real it found.

A second agent built a one-page brief before any call. A third drafted the outreach emails. Drafts only. Nothing left the building without my approval.

The research that would have eaten my evenings happened while I slept. The enterprise people who did not fit got scored 1 and archived automatically with the reason noted.

And none of it mattered, because the thing the funnel fed still needed 4 hours of me per client.

The lesson I paid six months to learn: automation multiplies a business. It does not change what the business is. Multiply a time-for-money service by a great funnel and you get an efficient way to sell time you do not have.

Check the engine before you build the fuel line.

I write a weekly email about building with AI, including the failures. Link in the comments.