Stop Measuring AI Productivity. Start Measuring AI Outcomes.

Pressure to justify AI investment is growing quickly.

Lalena Kennedy via Canva

By C200 Member Farrah Lakhani

CFOs and boards across industries are asking the right question about AI: where is the return on investment? It is the right question, but most organizations are measuring in ways that do not reflect real business impact.

Enterprise AI investment is accelerating rapidly, and pressure to justify that investment is growing just as quickly. Yet only 14% of CFOs in a recent survey report a clear, measurable impact from their AI initiatives. A Duke University study of nearly 750 executives points to what researchers describe as a “productivity paradox”: companies report meaningful gains from AI, but when those claims are tested against revenue and employment data, the real-world impact is significantly smaller across every major industry.

There is a straightforward explanation for this gap; it is not that AI doesn’t work. It is that most organizations are measuring the wrong things at the wrong time and applying expectations that do not align with how value is created.

Stop Measuring AI Productivity Instead Of Outcomes

The most common mistake I see is treating time saved as the primary measure of AI success. Hours saved, tasks automated and cycle times reduced are the easiest data points to capture and present in a board update. They are also the least likely to reflect real business impact.

Time saved does not equal value created. Consider a finance team that deploys AI to accelerate month-end close. If a ten-day process is reduced to six days but the time saved is absorbed into meetings and email, nothing changes on the P&L. The productivity gain is real. The business impact is not. The same pattern appears in revenue operations, where AI tools reveal pipeline signals that sellers receive but do not act on, and in procurement, where contract analysis tools identify savings opportunities that are never captured because no one owns the follow-through.

Organizations generating real returns are not measuring what AI does, they are measuring what AI enables. When I led the deployment of AI-powered forecasting tools across a large global sales organization, we did not measure success by the hours analysts saved from manual model-building. We focused on forecast accuracy, the gap between predicted and actual quarterly revenue, the downstream impact on capital allocation, and whether leadership made better decisions as a result.

The value came from improved decision quality. These metrics are harder to define upfront, require a baseline before deployment and take time to emerge. Many programs skip one or more of these steps, which is why productivity metrics look strong while P&L impact remains limited.

The Portfolio Problem

Gartner made this point directly at its Finance Symposium earlier this year: CFOs are misjudging AI investments by treating them as a single ROI problem rather than a portfolio of fundamentally different bets. A tool that reduces customer service volume has a very different economic profile than a model that improves how a sales leader reads their pipeline, which differs again from an agentic system redesigning an end-to-end workflow. Forcing all three through the same ROI framework, payback expectations, and cost-benefit logic produces outcomes that are often misleading.

In practice, a well-sequenced AI portfolio inside a large enterprise tends to include three distinct categories of investment:

The first is short-duration productivity tooling: AI-assisted drafting, automated data entry, meeting summarization and contract review. These show measurable adoption and efficiency gains within weeks and are relatively low cost to deploy. They build organizational confidence and create early proof points for continued investment.

The second category is process improvement: tools designed to change how core workflows operate, such as AI-enabled revenue leakage detection in finance, dynamic pricing models in commercial operations or intelligent billing platforms that proactively identify and resolve reconciliation gaps before they compound. These typically take six to twelve months to change behavior enough to show up in financial impact.

In one case, a finance team I worked with deployed an AI model to identify billing discrepancies and contract compliance gaps across thousands of customer accounts. The model identified issues within weeks, but turning those findings into revenue required redesigning escalation and collections workflows, aligning finance and sales, and several months of change management before the impact reached the P&L. The tool worked from day one. The outcome required everything around it to change as well.

The third category is transformation: redesigning how the business diagnoses performance, allocates resources or enters markets using AI-driven analysis that was not previously possible at speed or scale. This work rarely appears in quarterly revenue. It shows up later in the quality of planning decisions and the speed of response to market shifts.

If all three categories are measured against the same quarterly ROI standard, two of the three will be cut before they deliver returns. Treating AI as a portfolio helps leadership stay invested in the initiatives that take longer but ultimately matter most.

What Boards Should Actually Be Asking

Most board-level AI conversations fall into one of two failure modes. Either the board accepts a presentation full of adoption metrics and usage statistics without pushing on business impact, or it demands immediate P&L proof from initiatives that are not designed to deliver measurable impact within a single quarter.

The questions that produce useful information are different. Before any AI initiative is funded, the board should understand which business metric it is expected to move, what the current baseline is, and who is accountable for the outcome. Not which team owns the tool, but which leader’s performance is tied to whether it delivers.

During deployment, leading indicators that predict eventual impact should be reviewed alongside lagging ones. For a revenue leakage program, that might include the volume of identified discrepancies and the percentage resolved before the quarterly close, well before the impact appears in reported revenue. For a forecasting tool, it might include how often sales leaders override the model and whether those overrides improve or reduce accuracy. These signals show whether a program is on track long before final results are visible.

Boards also need to hold management accountable for measurement infrastructure before deployment, not after. ROI cannot be assessed if success was never clearly defined at the outset. This sounds straightforward, yet it is consistently overlooked.

The Right Timeline

CFOs, boards and the broader leadership team need to stop asking about this year’s returns and start focusing on how today’s investments will translate into value over time. Most enterprises see meaningful ROI from AI within two to four years, significantly longer than the seven to twelve months expected for most technology investments.

This does not mean we should accept slow performance or give programs a pass on accountability. We must build a measurement framework that distinguishes between what should deliver improvement now, what is building toward impact over the next year, and what is creating long-term advantage. Each category requires its own success metrics and its own board-level discussion.

The companies that pull ahead over the next three years will not necessarily be those spending the most on AI. They will be the ones measuring it like a capital investment, with clear baselines, outcome-based metrics, differentiated timelines by use case, and accountability built in from day one rather than added later when the board asks for proof.

C200 member Farrah Lakhani is a C-suite growth and strategy executive with more than 20 years leading AI-enabled transformation across global technology and financial services businesses. Farrah brings a rare combination of technology leadership, company-building experience, and deep financial discipline. She currently serves as Managing Director of Strategy, Planning & Operations at Uber, overseeing B2B enterprise-wide revenue strategy. Previously, she served as Global Head of Sales Strategy & Operations at Amazon Web Services. Earlier in her career, Farrah spent a decade as a Financial Services investment banker at J.P. Morgan, serving as a trusted advisor to CEOs and boards on M&A, capital strategy, and enterprise growth.