AI is technology, not a product
2026-05-27
In my first week consulting on a public-sector AI program, I collected a dozen different definitions of “AI” — just by asking people what they wanted.
That’s not the surprising part. The surprising part is how fluidly people swap those definitions inside a single conversation, at light speed. We’ve gone past AI-washing to AI-infusing: every vendor offers the definitive definition and appoints itself the authority on what’s in the club.
Some of the things “AI” meant that week:
- a chatbot like Claude or ChatGPT
- a vendor’s “AI-powered” dashboard
- software making real funding or policy decisions — human in the loop, or not
- an email tool that summarizes your inbox
So what is an “AI strategy” a strategy for? If everyone’s definition only partly overlaps, the word stops doing any work. With apologies to Orwell: in a world where everything is equally AI, are some things more equally AI than others?
Carl Sagan made the same point about ambiguous words — that agreement can just paper over real disagreement. He says it better than I will.
Here’s the framing I’ve landed on:
AI is technology, not a product. Applications have use cases that happen to leverage AI.
Sounds like a small distinction. It isn’t — and once you see it, you can’t unsee it.
Has anyone written an “internet strategy” in 20 years? We have e-commerce strategies, remote-work strategies, support strategies. The internet runs through all of them; it isn’t one of them. Same with electricity: no one appoints a Director of Electricity or catalogs “electricity use cases.” The technology is upstream of the decisions.
AI sits closer to that layer than to a product line. So the question stops being “what’s our AI strategy?” and becomes “should we automate this workflow? augment this role? surface insights from this dataset?” — with AI as one option alongside plain software, integration, process redesign, or deciding not to automate at all.
What changes when you take that seriously:
- You stop asking vendors “do you have AI?” — like asking “do you have software?” You ask what use case they solve, and what AI actually does versus the rest of the system.
- You govern around use-case risk and total cost, not around “the AI program” as one thing.
- You stop rewarding the blandest RFP language and start asking: what specific decision does this help us make?
Labels are easier than working definitions. They’re a fine starting place — but the real work is applying them to a real organization, with real people, process, and goals.
I’ve watched this play out in organizations from a dozen people to tens of thousands. I’d rather hear I’ve got it wrong now than agree-by-default and find out later.
Where are you seeing this? How are you handling it?