Fractional AI operations & deployment
AI projects rarely die on the technology. They die because no one — not the team, not the exec who funded it — can see clearly enough to fix the thing or trust it. I give you that clarity, in your language, and make it stick.
Eight years making a billion-dollar distributor’s automation clear enough that everyone from the warehouse floor to the CEO could trust it and act on it. Now I do it for you.
The gap
Four ways the fog shows up. One underlying problem.
The project can’t be finished because “finished” was never defined. Every review ends with “it’s close,” and close never converges. There’s no line the work is trying to cross.
There’s no instrumentation, so when you change something, you can’t tell whether it helped or hurt. So you stop changing it. The team settles on “it works well enough,” and that’s the moment it stops getting better.
It works on the clean examples from the demo and falls over on the actual mess your business runs on. The 5% of weird cases turns out to be where all the value, and all the risk, lives.
The team whose work it touches was handed a tool they didn’t ask for, don’t trust, and quietly route around. The thing technically ships and practically dies.
Notice what all four produce: a system no one can safely change. Every tweak is a gamble with no scoreboard, so the team stops tweaking, and “good enough” slowly rots as your data, your customers, and your business move on without it. A project you can see gets better every week. A foggy one just ages.
Every one of those is an understanding gap. Somebody can’t see clearly enough to act. So the thing I sell isn’t a model, a deck, or a dashboard. It’s understanding: a precise read on what’s wrong, proof that it’s the problem worth fixing, and that clarity delivered into the right heads in language they can act on.
What I am
For eight years, I owned the reporting on a billion-dollar distributor’s largest asset — an automated system that made most of its decisions with no human in the loop. My job was to make that machine legible: to see what it was really doing, catch where it drifted, and explain it plainly enough that everyone who depended on it could trust it and act. The understanding was the product; the inventory was just where you could see the work.
Finding the real structure in the noise is half of it. The other half — the half most people skip — is getting that structure into other people’s heads. I speak the languages it sits between — SQL for the data, Excel for the money, code for the system, plain narrative for the room — so the same understanding reaches the engineer, the operator, and the executive in the words each already works in. Knowing what’s true isn’t enough; it has to be understood well enough to act on.
Engagements
Each one leads into the next. Most start small, with a clear decision at the end.
I find what’s actually broken and prove it’s the problem that matters. Together we decide: fix, rebuild, or kill. What you keep isn’t a document, it’s understanding. You’ll see exactly what’s wrong and why, explained in your language, clear enough to bring your whole team along.
Take one process. I build the monitoring, the guardrails, and just enough tooling to make it run, then manage the people through the change so it lasts after I leave — because they understand it, not just tolerate it.
Embedded ownership of an AI initiative, a day or two a week. The judgment of a senior hire who keeps it clear, trusted, and on course — scoped to what you actually need.
01 surfaces the work → 02 fixes it → 03 keeps it alive.
Proof
I’m not the most AI-native person in the room. I’m the one who’s done the part that decides whether it works.
95% of buying was automated. 100% of outcomes were managed.
For eight years I built the instrumentation, exception management, and reporting that let leadership see, tweak, and trust it. It’s AI observability, before it had the name.
Steven is great at creating frameworks and processes that operationalize really complicated business problems, particularly when trade-offs are involved.
Former manager, McMaster-Carr
I taught 100+ people to read the data themselves.
The reporting only mattered because people understood it. I trained execs and analysts to see what the numbers meant, and wrote the documents that let leadership move faster — because, for once, they understood the call they were making.
Steven’s ability to dive into complex topics and communicate clearly to a broad audience results in strong decisions.
Performance review, McMaster-Carr
A shipped AI product, controlled.
I built a multi-model pipeline that layered deterministic business logic over the model output, keeping it consistent and trustworthy instead of erratic.
Steven translated the vision into a working product, with technical skill, business judgment, and a practical sense of what was worth building. Not just a tech lead — a strategic partner.
Reena, founder, StreetSmart
Two FTE, automated.
I inherited a process three departments each ran by gut. I pulled the decision logic out of people’s heads, encoded it, got leadership buy-in, and automated it.
He’s who you want on a messy problem with a lot of nuance — heads down, leading the team to build what I tell people we’re building.
Former manager, McMaster-Carr
The company’s biggest bets, modeled.
Briefing leadership on strategy and capital allocation, I built the model that cleared millions in dead inventory the profit-maximizing way, freeing capital to keep the business growing in its existing footprint. The same instinct now models capex and cash flow for founders building from scratch.
The financial modeling forced us to think about cash flow in concrete terms — where to invest, how to phase the capex for the build-out.
Founder, finalREV
“A bulldozer for internet bureaucracy.”
Founder, finalREVHow I work
I talk to everyone involved, find the real structure, build just enough to make decisions legible, and get people bought in to run it. I write code, but the code is rarely the hard part: it's getting everyone to see it.
— Steven
Start here
That’s usually where the real work is. Send me a line and we’ll figure out if there’s something worth doing.