Two transformations, one pattern
I've now led AI and data transformation inside two PE-backed companies. One grew from $120M to $1.2B during the pandemic era of healthcare staffing. The other shipped more than twenty production AI applications in a single year — with full adoption, not shelfware.
Different companies, different eras, different boards. But the longer I sat in those rooms, the more I noticed that the thing I was actually building wasn't AI products. It was enterprise value. And the things threatening that value were almost never on the technology roadmap.
The risk that never makes the board deck
Every PE-backed and owner-led company I've worked in or around carries two quiet risks.
The first is the legacy system. Not because it's old — old systems run profitable companies every day — but because of what surrounds it: manual reconciliation, tribal workarounds, and operating logic — how work gets priced, processed, scheduled, and handled when something breaks — that lives in habit instead of systems.
The second is sharper: the single point of knowledge. The one person who understands how the work actually moves. How it's priced. How it's processed. Why the system behaves the way it does, and where the bodies are buried in the data. Owners know exactly who this person is. So do diligence teams. When a buyer finds that the company's operating logic lives in one head, the discount is real, even when it's unspoken.
Neither risk shows up as a line item. Both show up at exit.
And both have the same day-to-day symptom: everything important moves at the speed of one person's calendar.
What actually worked
The transformations that worked never started with a platform decision or a hiring plan. They started smaller and more deliberately than anyone expected.
Learn from the people who know the business first. The expertise inside a twenty-year-old company is an asset, not an obstacle. The fastest path to working AI runs through the people who already understand the work.
Complement the legacy, don't fight it. Not rip and replace — rethink and extend. We built an AI-native layer alongside the systems that ran the business, and the business never stopped running.
Use the right kind of work for each step. Some steps want deterministic rules. Some want heuristics. Some want AI. Some want a person. The workflows that actually elevated the team combined all four deliberately — automation where it fit, judgment where it mattered — instead of automating for its own sake.
Turn transaction data into an operating asset before it becomes a diligence problem. Years of scattered transaction history is one or the other. The master data work we did with AI in weeks would have taken a traditional team months — and it became the foundation for better pricing, faster processing, and reporting people actually trust.
Prove value small, then scale. Every workstream that succeeded started as a small proof point with an internal owner. Every initiative that started as a multi-year program died as one.
Speed is the result owners feel
Here's what surprised me most: the wins that showed up in the numbers were really wins about tempo.
When reconciliation and reporting run themselves, decisions stop waiting on spreadsheets. When candidate response goes from hours to minutes, growth stops waiting on capacity. When the team can resolve long-standing tech debt in weeks instead of quarters, the whole company starts executing at a different clock speed.
Recruiters got five to ten hours a week back. The innovation team multiplied its output thirty-fold. Response times collapsed. And capacity went up while SG&A didn't — growth without headcount, which is the version of growth that survives an exit model.
Speed compounds. A company that answers questions about its own business in minutes, and acts on the answers in days, is structurally different from one that takes weeks for either. Buyers can feel that difference in diligence. Operators feel it every Monday.
Why I do this from the outside now
Here's the part I had to be honest with myself about.
I did that work with a C-suite title, a budget, and years of runway. For a long time I assumed that's what it took. I no longer believe that — and the math backs me up. Most mid-market companies cannot justify a full-time executive hire for this, and the ones that try often buy a new single point of knowledge with a bigger salary. The seat was never the ingredient. The method was.
The method transfers. An external advisor can learn from your people, build the proof points, structure the data, and — this is the part that matters — leave the capability inside your walls. Every workstream has an internal owner. Knowledge transfer is built into everything. Success is your team doing this work without me.
That's why I started GITLAI. External and fractional by design. No seat on the org chart, and no interest in one.
Where I'd start
If you own or back a mid-market company, the first questions aren't about AI at all:
Where does our operating logic actually live — how we price, how we process, how we handle exceptions — in systems, or in someone's head? How fast can we answer a basic question about our own business, and how fast can we act on the answer? And if our most knowledgeable operator left in ninety days, what breaks?
If those questions are uncomfortable, that discomfort is the roadmap. The companies that answer them now — before the exit process forces the issue — are building the kind of value that survives a change of ownership.
That's the work. It starts small, it runs through your people, and what it leaves behind is yours.