Why Enterprise AI Programs Become Unprofitable in Year Two
Enterprise AI investment has crossed a visibility threshold. For a typical ten-thousand-person enterprise, the annual spend on agentic AI now sits at a scale the CFO can no longer treat as an innovation line item. The question has shifted from whether to invest to whether anyone can prove the investment is working. Most enterprises cannot.
This is not a finance failure. It is a structural mismatch between how AI systems behave in production and how enterprises account for technology. Traditional software carries a build cost and a maintenance cost that decays. Agentic systems carry a third cost: continuous, variable, and compounding, incurred every month the system runs. This briefing names it the Operating Tax.
The briefing sets out the three layers of AI economics, locates the Inversion Point at which an agent’s cost per outcome exceeds its value per outcome, and gives finance and technology leaders the instruments to see the inversion coming before the year-two budget review finds it for them.
