Skip to content
GAT.
Case studies

See the shapebefore you pay.

Three worked examples, one per step of the engagement. baxie is real and running in beta today. The other two are composites built from real patterns, with illustrative numbers, so you can see how an engagement runs without a client's name on it.

Book 30 minutes with Edwin

Every engagement starts with the AI Program Diagnostic. $2,500, two weeks, credited in full if you build.

01 · The examples

One per step.One of them is real.

GAT.

Step 01 · The plan · Composite

Industrial distributor

A new COO inherits three stalled AI pilots and a board deadline. The diagnostic maps the work, measures the baselines, and hands back a 90-day plan with a kill list.

30 hrs/wkRFQ re-entry, baseline
GAT.

Step 02 · The build · Real

baxie

The construction SaaS I built and run in beta. Vision extraction on plan sets, multi-model routing, and a review gate enforced in code. The page also says what it cannot do yet.

In betaNot a composite
GAT.

Step 03 · Keep me on · Composite

Healthcare SaaS

The AI lead quits and three pilots drift. I triage the portfolio, ship the claims workflow that mattered, and stay until a permanent owner signs.

40% → 58%Tickets resolved without a human
02 · How to read them

Composites.Labeled as such.

The two client examples are composites. The company and the numbers are illustrative. The shape of the work is real: the workflow map, the baselines, the build types, the kill list, and who owned what.

baxie is the exception. It is my own product, it is live, and its page lists the limits I will not paper over. If you want to check how I describe a system when the numbers are not flattering, start there.

Your plan will not look like either composite. It starts with your system data and interviews with the people who do the work.

GAT.

Bring your own workflows. I'll tell you where to start.

30 minVirtual call

If there is a fit, I will scope the diagnostic. If there is not, you leave with a sharper read on your next move.

Book 30 minutes with Edwin

Not ready to talk? Score your AI program first. Twelve questions, about four minutes, scored breakdown by email.