AI ROI Has a Timing Problem.

We have spent months telling leaders that AI must be measured.

That remains true.

But a different problem is beginning to show up: companies are measuring the right outcome at the wrong moment.

They launch an AI pilot, wait a few weeks, and ask whether revenue or profit has moved. When it has not, they question the investment. Or they do the opposite: they report logins, messages, and hours saved for months without ever connecting any of it to a financial result.

The first approach can kill a valuable workflow before it has had time to change the business. The second can keep an unproductive pilot alive long after it should have been redesigned or stopped. This week’s clearest signal came from Infosys. Its new survey of more than 1,000 senior executives found that 75% believe AI’s total value to their organization has been net positive. Yet 67% still struggle to measure AI ROI, and 74% say pressure to prove short-term ROI is inhibiting more transformative, long-term initiatives.

That changes the leadership question.

Instead of asking only, “Has AI reached the P&L yet?” ask, “What evidence should exist at this stage—and is the workflow producing it?”

Here are the numbers—and what they mean for professionals trying to move AI from early activity to measurable return.

The numbers worth knowing this week

The pattern that matters

AI value does not arrive as one event. It moves through a chain.

First, people have to use the new workflow correctly. Then the workflow has to perform better than the process it replaced. Only after that improvement reaches customers, capacity, conversion, retention, cost, or risk can leadership defend a financial return.

Those are three different tests:

●​ Adoption proof: Are the right people using the workflow consistently, following review rules, and correcting weak outputs? ●​ Operational proof: Is the work becoming faster, more complete, more accurate, less expensive, or easier to scale? ●​ Financial proof: Did that operating improvement ultimately change revenue, profitability, retention, cost to serve, or risk exposure after the full cost of the AI system is included?

The mistake happens at both ends of the timeline.

If leaders demand the final P&L result before adoption and operating performance have stabilized, they may stop a promising workflow too early. If teams continue reporting seats, messages, and hours saved after months of deployment, they may keep an ineffective system alive too long.

Every initiative should have a financial destination before it launches. But each stage needs its own evidence, and the review schedule should reflect the business process being changed. A customer-service workflow may show cost and quality movement quickly. A recruiting, retention, or relationship-building workflow may need months before the financial result is visible.

This is not a case for waiting indefinitely. It is a case for agreeing in advance on what should change first, what should change next, and when the final financial test becomes fair.

The EMDUIT take — for the professional putting AI to work

Consider a real estate brokerage deploying an AI recruiting workflow.

If leadership measures only signed recruits in the first month, it may miss the work the system must improve before a recruit ever joins. First, the brokerage needs adoption proof. Are the right prospects being identified? Is approved outreach being completed? Are responses being handled quickly? Is the broker receiving qualified conversations instead of more administrative work? Are employees using the workflow consistently, or quietly returning to spreadsheets and manual follow-up?

Then the brokerage needs operational proof. Are more prospects entering real conversations? Are fewer opportunities going cold? Are appointments increasing? Is follow-up becoming more consistent without losing the broker’s voice, judgment, or compliance standards? How much time and cost does the complete recruiting workflow now require?

Only then do the financial measures have context: recruits who join, activate, remain with the brokerage, and ultimately contribute to company dollar—compared with the platform, implementation, maintenance, oversight, and human time required to produce that result.

If the brokerage demands signed recruits before the workflow has created consistent conversations, it may stop too soon. If the team is still celebrating messages sent six months later, it has waited too long.

This is why accountability and patience are not opposites.

Before implementation, document the current baseline and define the financial destination. During deployment, report the leading signals. After the workflow stabilizes, test the operational change. At the agreed review point, reconcile the result with the full cost. The organizations that gain ROI from AI will know what evidence should exist at each stage—and refuse to confuse one stage with another.

Which AI initiative in your organization is being judged on the wrong part of its value timeline?

If you are a licensed professional or business leader, share the workflow that needs a clearer path from adoption to operating performance to financial return.