What we'd do in your first 30 days

A friend called last month. He runs a 22-person business. “We’re drowning in customer emails. Should I hire someone, or should I use AI?”

I told him I didn’t know yet.

That’s not modesty. It’s the right answer. You can’t recommend the tool until you understand the work — and a 30-day pilot is the smallest commitment that gets us both there honestly. Here’s what those four weeks look like.

Week 1 — Diagnose. We talk to the people doing the work, not the people describing it. We ask what they did yesterday, not what they “generally” do. The first answer is fiction. The second is data. By Friday you have a one-page audit: where hours leak, what they cost, what’s worth solving in 30 days.

Week 2 — Design. We map the workflow on paper before we touch a tool. Trigger → action → handoff → output. Half the workflows we’re asked to automate don’t need AI — they need a decision. We’ll tell you which is which. By Friday: a workflow blueprint your team can read in one sitting.

Week 3 — Build. Hands on keyboards. Could be a custom GPT, a Make.com scenario, a Claude skill, or off-the-shelf. We pick what fits your stack, not what’s trendy. The last days of week 3 are dedicated to making sure the system survives without us — runbook, training, the “what to do when it breaks” script.

Week 4 — Measure. Real numbers. Hours saved per week. Errors reduced. Time-to-completion delta. If the workflow didn’t stick, we name why — usually it’s process, not tool. Most consultants disappear here. We don’t.

Where this fails. Two of the last five workflows we’d pitch this for didn’t need AI. Two of the others would have failed without a clear owner on the client side. The 30 days only works if both halves show up.

If you’ve got a workflow eating more hours than it should, the next step is a 15-minute call. Not a sales pitch — we ask you what’s going on, you ask us if we’re a fit. If we are, we quote a pilot. If we’re not, we’ll tell you who is.