And I'll tell you honestly whether you're ready for AI before someone sells you some.
One person maintains it. It pulls from three systems that don't talk to each other, it gets rebuilt by hand every month, and if that person left tomorrow you'd have a problem nobody wants to say out loud.
Meanwhile your board, your bank, or your best customer has started asking what you're doing with AI.
Here's what nobody selling you software will say: you can't automate a mess. You can only automate a process. Every AI project that fails in a mid-market company fails for the same boring reason — the data underneath it disagreed with itself, and nobody checked first.
You'll recognize yourself in most of these.
No full-time data person. IT keeps the lights on, but nobody owns the numbers.
An ERP or field system, a CRM, accounting, payroll, and a pile of vendor portals.
Every location or division ended up doing it slightly differently.
Days a month spent producing a report that should take four minutes.
And you genuinely can't tell whether it would work on your business or not.
I'll be in your office, not on a status call from three time zones away.
Manufacturers, distributors, contractors, energy services, multi-location operators, professional services firms. The industry changes. The problem doesn't.
Two to three weeks. Deliberately small, deliberately cheap — because it's how we both find out whether we should work together on anything bigger.
Every system you run, what's in it, how it connects, and where the same number lives in three places with three different values. Most owners have never seen this drawn. It's usually the most uncomfortable and most useful page in the engagement.
Eight dimensions: completeness, duplicate and matching problems, timeliness, where numbers come from, who can see what, whether "revenue" means the same thing in two departments, how well systems integrate, and what's documented. Standardized — so your score means something against other companies your size, not just against your own gut.
Not a wish list. Each one scored on value against feasibility, with a real effort estimate and a real dollar figure. At least one automation or AI play. At least one piece of basic reporting you should already have and don't.
Not a mockup. Something that runs: a reconciliation that finds actual discrepancies in your actual numbers, an automated version of that monthly spreadsheet, or a tool that pulls data out of the PDFs your team is currently retyping. You keep it whether or not we work together again.
The map, the score, the roadmap — with prices attached, so you can decide what happens next without sitting through another sales pitch.
Not hourly. Not a range. Not plus expenses.
And if you sit through the readout and don't think it was worth it, don't pay.
I can offer that because it hasn't happened. If it does, I'd rather find out than argue about an invoice.
All fixed-fee. You're buying an outcome, not my hours.
| Engagement | What it is | Price |
|---|---|---|
| Embedded build | I work inside your business for 3–6 weeks and ship the highest-value item from the audit. | $18–35k |
| Single fix | One specific thing, built and handed over. Two to four weeks. | $8–15k |
| Ongoing | One to two days a week, on site. I own your data function the way a full-time hire would, without the hire. | $6–9k/mo |
For context: a competent data engineer in Houston costs $130–170k plus benefits, plus six months to find, plus the risk they leave. This is the version where somebody senior shows up next week.
One person. The one who does the work, in the room, every time. No account manager, no bench being billed to you, no junior you've never met showing up on week two.
I'll ask what systems you run and where the manual work is. If the audit isn't the right thing for you, I'll say so on that call, and it costs you nothing.
Start a conversationOr call directly — [ PHONE ]
I take two audits at a time. That's a real constraint, not a closing line.