If your AI initiative succeeds — really succeeds, the model works, the adoption lands, the workflow changes — what stops?
It's a question most plans can't answer. There's a vivid description of what AI will do — accelerate this, automate that, surface the other thing — and a very thin description of what gets retired, redeployed, sunset, or shelved. The work the AI was supposed to replace is still on the books. The team that was supposed to be freed up is still doing what it did before, plus the new tooling. The cost base goes up. Output sometimes goes up too, which keeps everyone polite about the ROI conversation.
Addition is not strategy. Addition is wishful thinking with a budget line. Real strategy names, in advance, the thing that will be smaller or gone if the project works — a product line, a function, a vendor relationship, a category of meeting. The naming is the hard part because it has political costs the addition story does not. The addition story can be supported by everyone whose job is the thing being augmented. The subtraction story costs someone something specific.
This is why so many initiatives produce confidently more output and confusingly flat margins. The work the AI was supposed to displace is still being done, because nobody said out loud which work it was. The AI bolts on rather than replaces, and the org ends up more expensive, more complicated, and not noticeably more profitable.
"AI strategy" without subtraction is just a higher cost base with extra meetings.