AI & ML strategy

Where to invest, what to ignore, and how to manage the risk — judgment over hype, from operators with no product to sell you.

Every mid-market executive is getting the same two messages right now. From the market: move on AI or be left behind. From experience: the last three technology waves came with the same warning, and the companies that panicked bought the most and gained the least.

Both messages contain truth. Sorting out which applies to your business, this year, is a judgment call — and judgment is the part the vendors can't sell you, because every one of them arrives with the conclusion pre-written.

A board or owner asking for "our AI strategy" and a leadership team that suspects the honest first draft is thinner than the deck implies. A pilot that went well in the demo and sideways in production. A budget cycle where three departments each requested AI tooling and nobody can compare the claims. Or a PE sponsor who sees AI in the value-creation plan and wants to know which parts are real.

And increasingly: agents — software that acts, not just drafts — arriving faster than the rules for what it's allowed to touch.

An executive who has evaluated and deployed these technologies as an operator — not a lab, not a reseller — works through your business, not the technology landscape: where your margin actually comes from, which workflows have the volume and error-tolerance where today's AI genuinely pays, and which headline use cases would be expensive theater in your context.

The risk side gets equal weight: data boundaries, vendor exposure, the governance an insurer or customer will eventually ask about, and the human question we wrote a whitepaper about — how to adopt these tools without outsourcing your judgment to them.

A short, funded, sequenced plan: the two or three investments worth making now, with owners and success measures; the list of things to deliberately skip this year, written down so the decision holds; and a governance baseline sized for a mid-market company — a page of policy, not a bureaucracy — so pilots can move fast without creating the incident that ends the program.

Not a research report on "the state of AI." Not a vendor bake-off with a predetermined winner. And not a promise that AI transforms everything — in most mid-market businesses the honest near-term answer is narrow, boring, and profitable, and we consider defending that answer part of the job.

For the thinking behind the approach, read AI Ethics & Human Cognition — our whitepaper on adopting AI without surrendering judgment — or the shorter briefing on agentic AI from this spring's archive.

When the board asks for the strategy, the first step is an hour, not a workshop: start the conversation.