I help enterprises make AI actually stick: the technical depth to build it, and the pattern recognition to know what gets adopted in the messy reality of a large organization.
Twenty-plus years in data. Co-founded and bootstrapped Aptitive, a data consulting firm delivering data solutions to midsize enterprises, from 2013 to a successful exit in 2021. Hands-on with LLMs since early 2022, before ChatGPT, and every single day since, across Anthropic, OpenAI's Codex platform, Gemini, and open source models. Each has its own behaviors and quirks, and there's no substitute for using them daily to understand them.
That daily use is also how I can tell who can actually build with this, which matters because hiring in AI is hard right now: a strong resume from two years ago says little about whether someone can build with agents today. So I help clients find and hire the right people. And I spot the wrong path early. The detour that would have eaten a week, a quarter, or the whole project, I can usually unstick or simplify, or get the executives aligned on one goal.
What this looks like in practice
Every engagement is different, but they tend to touch three levels of the organization, and the playbook for each has carried from one client to the next.
Small, high-powered AI-native teams
Stand up a small team inside the organization that lives on the frontier and works like an internal consultancy: it takes on problems from across the business and teaches the rest of the org as it goes. Usually that's a mix of early-career engineers new to the workforce and a few senior hires, and I help find both. Then I get them productive fast on how agents and harnesses actually behave, where they fail, data flywheels that hold up, and the human-feedback loops that make systems better over time.
I stay hands-on the whole way: mentoring, showing what works and what doesn't, and earning the team's trust by building alongside them. Their own talent does the rest; I just speed it up. It's the same thing that worked at Aptitive: small teams of curious, driven people from a diverse set of backgrounds just work when done right.
You get a team that ships without me and stays on the frontier as the technology moves, plus a clear hiring bar for who to bring on next.
Business rules that write themselves down
Put Claude Enterprise (or the equivalent) in the hands of business teams across functions and let them automate the tedious parts of their jobs. The knowledge that lived in someone's head or in an Excel file becomes written-down business rules and reusable skills, including the exceptions and judgment calls that never make it into a requirements doc. That's how new people get onboarded, and it's a better source of truth for downstream AI, analytics, and automation than any spec.
It also gives business users a way to build and own their own apps, dashboards, and analyses, the kind that never got budget before, which is why they lived in Excel. Some are single-use: a forecast that matters today and never again. The patterns and skills behind them stick and get reused.
You get a business that intuitively knows how to use these tools, faster onboarding, and a library of how-the-business-works that keeps growing.
Adoption that compounds across the organization
Let the business drive. Leadership clears the path, teams use the tools on their real work, and feedback comes from two places: the teams themselves, and usage data. For the data side I build the analytics layer over enterprise platform adoption that shows leadership which teams are using it, for what kinds of work, and what's growing, without storing or reading a single prompt or response. Each team that gets good at this makes the next one faster, and the value compounds.
You get adoption driven by the business instead of pushed at it, C-suite visibility into real usage with privacy intact, and a clear view of where to invest next.
How I work
Data-centric. Business rules, usage, feedback, and the solutions themselves are all treated as data. That's why each solution becomes the starting point for the next one, and why these patterns have kept scaling since agentic workloads were new, through every model release since.
The people who know the domain stay in control. AI should make them more productive and more engaged, not sidelined. Meet the organization where it is, not where a vendor thinks it should be.
Think platforms, not models. The model matters less every quarter. What matters is that the providers, Claude, OpenAI's enterprise offerings, and others, are building out into something closer to clouds, where new kinds of workloads run directly inside them. That's the useful frame for choosing one and planning around it.
Simple-yet-scalable beats clever. In agent design and in org design.
Work together
Fractional engagements under my own firm, Intersect Next, mostly with enterprises. Format follows the problem: strategy sessions, hands-on building, or some mix.
Fractional AI / technical leadership
Architecture, hiring, and mentorship for the team doing the work. Closing the gap between the vision on the slide and what ships.
Strategy & advisory
Unstick a stalled initiative, align executives on one goal, sort what's real from what's noise.
Implementation partnership
Build it together, transfer the knowledge, leave the team stronger than when I showed up.