What is AI actually returning?

AI spend has become easy to see and hard to justify. As generative AI moves from pilots into daily operations, budgets are shifting from flat licenses to consumption pricing, and the line item grows with every workflow it touches. The evidence for value, meanwhile, arrives as surveys: useful for sentiment, silent on mechanics. This piece sets out how we connect an individual AI lever to a measurable Return on Transformation Investment, and where to begin.

Why is the value of AI so hard to pin down?

Cost lands in one place; value scatters. The invoice sits with IT or a central AI budget, while the benefit surfaces as shorter cycle times in operations, fewer touches in finance and better conversion in sales. Each function sees a fragment, and no one owns the sum. The leading surveys compound the problem. They report what respondents believe AI delivers, without showing which practices produced it or where it lands in the financial statements.

Where does AI value actually show up?

In three places. The income statement, through revenue, cost of goods and operating expense. The balance sheet, through working capital, inventory and asset intensity. And risk, through quality escapes, compliance exposure and resilience. A benefit that cannot be routed to one of these lines is a claim, not a return. Mapping each use case to its line before funding it is the most clarifying discipline a leadership team can adopt.

What is a Return on Transformation Investment?

RTI reads AI value from the bottom up. It starts with the lever, the specific change in how work gets done, measures it, then rolls it up through the process, the function and the enterprise. It reads levers independently and collectively at the same time. Two automations that each look marginal can compound when they share data or remove the same handoff; two that each look strong can cannibalize one another. RTI makes those interactions visible before capital is committed, and the same lens travels across every company in a private equity portfolio.

The useful question is where AI creates value, how much, and at what cost per successful outcome.

Where should a company start?

With the question leadership is already asking. Our playbook is organized around six of the most common entry points, among them “we need to reduce what we spend on AI” and “what value are we really getting?” Each opens a different path through the same framework, so the work begins where the pressure is and still arrives at a single enterprise view.

How do you keep the gains?

Give every lever an owner, a baseline and a target on a named line of the P&L or balance sheet. Re-read the portfolio each quarter, retiring levers that no longer earn their consumption and investing behind the ones that compound. Value measured this way tends to stay captured.