Worked examples.
Worked examples: how I'd think through real situations in industries I know. If you want to see how I reason before we ever talk, start here.
A 65-person manufacturer losing margin to rework.
The spreadsheet says it's a QA problem. Usually it's a data problem — the defect clusters stay hidden until the numbers live in one place.
First thing I'd do: pull the rework logs together and find the top three defect sources. That's a two-week exercise — a tight first pass before any talk of automation.
A boutique hotel group losing bookings to slow replies.
Guest messages pile up across booking platforms, email, and text — and the slow replies cost bookings nobody ever counts. The data to fix it is already sitting in the PMS.
First thing I'd do: route every channel into one inbox and draft replies to the routine stuff, so staff only touch the messages that need a human. Then look at what the reviews keep complaining about — that's free market research.
A clinic where the front desk is buried in intake paperwork.
Intake is the most automatable part of a practice and the least automated. Paper forms, re-keyed data, eligibility checks by hand — it's hours a day of work nobody chose.
First thing I'd do: measure how long intake actually takes per patient, then pre-fill from existing records and flag what's missing before the patient arrives. Everything stays inside the practice's systems. HIPAA is the constraint, built in from the start.
A grower-packer sitting on years of yield data nobody looks at.
Most operations are already collecting the data — yields, grades, weather, shipments. It just rarely gets looked at together. The first win is usually making it visible — before building anything exotic.
First thing I'd do: pull yield history, weather, and order data into one place and forecast harvest and shipment needs against it. Packing-line vision comes later, once the data story is clear.
A GC spending Sundays reading submittals.
Construction runs on documents — estimates, bids, submittals, RFIs, change orders. The work is plentiful — the hours to process the paperwork around it are the scarce part.
First thing I'd do: look at estimating, because every bid starting from scratch instead of from your own history is margin leaking out. An assistant trained on past bids gets the estimator to the number faster — they still own it.
A property manager whose inbox is the bottleneck.
Property management is a communications business wearing a maintenance business's clothes. Tenant messages, maintenance triage, rent follow-up — and the portfolio keeps growing faster than the staff.
First thing I'd do: sort the tenant inbox automatically — emergencies to the top, routine questions drafted for one-click replies. Then look at maintenance triage: categorize, estimate urgency, route to the right vendor, coordinator approves.
Let's find your business's primary bottleneck.
The first step is a workshop: $5,000, two weeks, one clear answer about where AI pays for your business. You get a straight read either way: where to invest, and where to hold.