Why it matters now
I see the same pattern in many leadership teams. AI spending is broad, usage is growing, and nobody can point to the line in the accounts where it shows up. That is rarely a technology problem. It usually means the work was never tied to a business measure, or the full cost was never counted. It is fixable, and it starts with deciding what success means before the next project begins.
- Only about 25% of AI initiatives have delivered their expected return on investment.IBM Institute for Business Value, 2025 CEO Study ↗
- When CFOs measure the return on generative AI, they most often track customer experience (78%), margins (75%) and operating costs (60%).PYMNTS, CFO generative AI ROI survey, 2025 ↗
- Klarna’s CEO said cost had been “too predominant” a factor in its AI push and that quality fell. The company began hiring people again.Fortune, Klarna and AI, 2025 ↗
Common mistakes
- Counting usage, seats or projected savings instead of results that show up in the accounts.
- Leaving out the full cost: integration, training, review, governance and maintenance.
- Funding many scattered pilots with no link to the work that creates most of the value.
- Treating headcount reduction as the main measure, and letting quality slip unnoticed.
- Setting only efficiency goals and never asking where AI could help the business grow.
What to do instead
- Set a baseline first. Before anything is deployed, record how the work performs today: time, cost, error rate, customer outcome. Without a baseline, any result is a guess.
- Name one business measure per initiative. Each project gets one profit, cost or customer number it is meant to move, and one owner who answers for it.
- Count the whole cost. Include models, integrations, review time, maintenance and the effort of changing how people work. The value is what is left after that.
- Separate early signals from results. Adoption and quality tell you early whether something is working. Margin, revenue and customer results confirm it later. Track both, and do not confuse them.
- Set growth goals as well as savings. Ask where AI could help you win or keep customers, not only where it could cut cost.
- Review value at the top. Put AI value on the executive agenda every quarter. Scale what shows results, and stop what does not.
Your first step this week
List every AI initiative you are paying for. Next to each, write the business measure it should move and its baseline today. Any line you cannot fill in is your first conversation.
Related guides
- How do we get from AI pilots to scale?
- How fast should we move on AI, and how much should we invest?
- How do I work with the board on AI?
All AI leadership principles and guides · 100 AI use cases by function