Someone has to own the AI program. That is the job I do. Ex-eBay Technical Program Manager, now shipping production AI as the founder of baxie.
Start with the AI Program Diagnostic: the plan, not the build. $2,500, credited in full if you build. Next opening October 6.
The diagnostic gets you the plan in two weeks. The build gets you the rest. Ninety days after kickoff, you have four things you can point at:
That is the whole job. Someone owns the program, the first system ships, your team takes it.
You have tried a few things. Something demoed well. Very little of it is doing real work.
AI programs fail the way engineering programs fail, and the failure has nothing to do with the models. The work itself takes minutes. The process around it takes weeks, and almost all of that time is waiting on handoffs nobody is tracking.
AI is everyone's priority and nobody's job. When a workflow produces a bad result, the answer is "the AI did it," and that answer does not survive a board meeting.
A model gets dropped onto the workflow as it is. The step it touches gets faster. The four handoffs around it stay, so the elapsed time barely moves.
Nobody wrote down the old number before the pilot started. Six months later nobody can say whether it worked, so it never gets funded or killed.
That is a program problem. It gets solved by someone who has run programs, not by buying another tool.
I map the workflows your AI work touches, from your system data and interviews with the people who run them. You get a 90-day plan ranked by the handoffs each line removes, with an owner, a baseline, and a build type on every line. See what the plan looks like.
The first lines of the plan, shipped into the stack your team already runs. No migrations. Agents start by recommending and earn more autonomy only when the evals prove it. Every build is measured against the baseline in the plan, so your CFO sees what it returned.
Training so your people run the systems, read the evals, and spot the next opportunity without me. The engagement is done when your team can take it, not when the code ships. If you want me to stay, there are two monthly options: keep it running, or keep building.
Before you askBuilds are priced per workflow shipped, not per hour: from $7,500 a workflow, two-workflow minimum. After the build, System Continuity keeps it running at 2 to 3 days a month, and the Fractional retainer keeps building from $12,000 a month for one day a week, up to three days. 14-day mutual exit from day one on every retainer.
Every engagement starts with the AI Program Diagnostic. $2,500 buys the plan. It does not build anything. Day 1 is a 60-minute kickoff. Days 2 through 9 I map your AI work. Day 10 you get the readout. Paid before kickoff, so there is no scoping back and forth and no change orders.
$2,500 flat. Credited in full.
What the $2,500 pays for: eight working days of my time, interviews with the people who run each workflow, and the numbers pulled from your own systems. On the retainer, eight days bills at over $20,000. What it protects: a two-workflow build is $15,000, and the plan makes sure that money goes to the right two.
Not included
Building, integrating, or training. That is step 2, the production build, priced per workflow from $7,500 with a two-workflow minimum. The price is fixed in writing at the readout. You see the plan before you decide whether to build any of it, and the plan is yours to take anywhere.
Four lines from a 90-day sequence for an 80-person services firm. Illustrative, not a client. A real plan runs 10 to 20 lines, and every line carries the same columns. Those columns are what the guarantee checks.
| # | Workflow | Owner | Baseline today | Build type | Handoffs removed |
|---|---|---|---|---|---|
| 01 | Weekly pipeline report Quick win | Sales ops lead | 3 hrs a week, pulled by hand from the CRM | Plain code | 1 |
| 02 | Invoice to purchase order match | AP manager | 20 min of work, 6 days elapsed, 4 handoffs | Agent, human approves | 3 |
| 03 | New client intake to kickoff | Director of operations | 11 days elapsed, 5 handoffs across 3 teams | Code plus agent | 3 |
| × | Website chatbot pilot Kill | Marketing lead | 40 chats a month, none tracked to a sale | Stop it | 0 |
Build types: plain code when a rule does the job, an agent when it takes judgment, human review when a mistake costs more than the time saved. Most lines turn out to be mostly code.
Ex-eBay TPM.
Founder of baxie.
Built for operators.
Technical Program Manager at eBay, reporting to a VP of Engineering. My peers were directors and engineering managers. My day job was getting roughly a dozen teams across Core Product and Core Technology to agree on trade-offs.
Owners, handoffs, risk, downstream impact. It is the same map an AI portfolio needs. Strategy firms draw it and cannot ship against it. Automation shops ship without drawing it. I do both.
What "fractional" means here: I do the work a full-time AI leader would do, at the fraction of time the work actually needs. Without the seat, the equity, or the 9-week search.
Not ready to talk
Twelve questions, about four minutes. You get the scored breakdown by email: where your program is exposed, which dependency is blocking the rest, and what to fix first. No call required.
Take the scorecardAn agency hands you a tool and leaves the program unowned. A hire takes nine weeks to find and six months to ramp. I own the program from week one, ship the systems, train your team, and hand it off.
I build. The plan exists so the build lands. The two are sold separately: the diagnostic is the plan, $2,500 flat. The build is scoped and priced in writing at the readout, and you decide after you have seen the plan.
The plan lands on day 10, and its quick win is something your team can ship that week. The first production build ships inside 90 days of kickoff.
A free audit usually runs on a survey and five or six interviews. A paid plan goes wider: I interview the people who run each workflow and pull the numbers from your systems. That depth is what earns the right to price the build, line by line. Every line carries an owner, a baseline, and a build type or you get the fee back, and if you build within 30 days the full $2,500 comes off. If you want a list of ideas, take the free audit. If you want a plan I'll put a refund behind, that is what $2,500 buys.
Typically 20 to 200 people. The size matters less than the situation: you own the AI decision, you have tried at least one use case, and nobody is running the program. The same situation shows up in professional services, distribution, healthcare software, and single-family offices. Program discipline travels across industries.
On your stack, under your controls. The shared knowledge base is the layer your team and your AI both read from. It is not a place your data leaves, and it is not a product I resell you.
The AI Program Diagnostic is $2,500 flat, paid before kickoff, and credited in full if you build within 30 days. Builds are priced per workflow shipped, from $7,500 a workflow with a two-workflow minimum. The Fractional retainer starts at $12,000 a month for one day a week. I do not bill hourly and I do not take vendor referral fees, so the number does not move once it is written.
If there is a fit, I will scope it. If there is not, you leave with a sharper read on your next move.
Book 30 minutes with EdwinNot ready to talk? Score your AI program first. Twelve questions, about four minutes, scored breakdown by email.