The Wrong Starting Point
It is Q4, which means enterprise AI budget reviews are in full swing. And in most companies, the conversation starts with a spreadsheet showing AI spending by category: compute, licenses, headcount, consulting. The CFO asks whether the spend is justified. The AI team scrambles to show ROI. Everyone leaves frustrated. This conversation has happened in every company we work with, and it is the wrong conversation to be having.
Questions That Actually Matter
- What decisions did AI change? Not what decisions AI could theoretically change. What decisions actually changed because of an AI system in production? If the answer is none, your AI program is a research lab, not a business initiative. Track this metric explicitly: decisions influenced, with dollar values attached where possible.
- What did we stop doing because of AI? AI that adds work without removing work is a net cost. The real ROI comes when AI enables you to eliminate a manual process, sunset a legacy system, or reduce headcount in a specific function. If you cannot point to something you stopped doing, AI is adding complexity, not reducing it.
- What is our time-to-production for AI use cases? Measure the elapsed time from approved use case to production deployment. If it is longer than six months, your bottleneck is not technology or budget. It is process, governance, or organizational readiness. Spending more money will not fix a process problem.
- What did we learn from failures? Every AI program has failed experiments. The question is whether those failures generated insights that improved subsequent efforts, or whether they were quietly shelved and forgotten. An AI program that fails and learns is healthy. One that fails and hides it is wasting money while also destroying the feedback loop that enables improvement.
Reframing the Budget Conversation
The AI budget should not be evaluated like a cost center or a capital project. It should be evaluated like an R&D portfolio. Some bets will pay off enormously. Some will fail. The success metric is not ROI on every dollar spent. It is whether the portfolio is generating compounding capability that makes the organization more competitive over time.
The Practical Move
Before your next budget review, build a one-page document that shows: decisions changed, processes eliminated, time-to-production trend, and lessons from failures. If you cannot fill that page, the problem is not the budget. The problem is that your AI program is not connected to business outcomes.