AI is generating excitement across organizations, but many companies still struggle to prove its real business impact. The core issue is that AI is often measured using activity metrics—like adoption or usage—instead of outcomes that matter, such as revenue growth, cost reduction, or efficiency gains.
To close this gap, organizations need a more disciplined approach to measuring AI ROI. This starts with clearly defining the business outcome an AI initiative is expected to influence, then establishing a baseline before deployment so improvements can be proven. From there, measurement should move beyond individual tasks to evaluate full workflow impact, where the true value compounds across processes.
It’s also critical to capture both direct value (like cost savings or increased revenue) and indirect value (such as improved user experience or faster decision-making), while tracking how AI performance improves over time. Unlike traditional investments, AI delivers compounding returns as it learns and scales, making ongoing measurement essential.
Few people in information systems and technology have seen as much excitement about a new technology as we see today. The internet changed everything, but not many recognized its potential early on. AI is different. It is transformative, and almost everyone, from CEOs to new interns, understands its potential.
The excitement about AI is real. But from a CIO’s perspective, there is a clear gap between that excitement and proven business results. Companies are spending more on AI every quarter, yet many still struggle to answer the question boards and CFOs are now asking directly: what is this actually delivering?
For most companies embracing digital transformation, the honest answer is that we’re not entirely sure.
This uncertainty can’t last. AI is a business investment, and like any investment, it must answer the question: What is the return on investment? We need to measure AI’s impact in terms of value creation, not just activity. If we do this, we can turn enthusiasm into real business value.
Why Measuring AI ROI Is So Challenging
Part of the challenge is structural. Traditional IT projects, like cloud migrations, have clear costs and benefits. They have a set scope, a baseline, and outcomes that can be measured after deployment. AI is different. It’s probabilistic, changes over time, and involves many teams. Its benefits often show up in hard-to-measure areas like time saved or process improvements, and these gains spread across workflows in ways that are hard to track.
This complexity is made worse by how most AI programs are managed. Often, teams skip setting clear baselines before starting, driven by excitement to move quickly. While speed is good, launching AI without measuring current performance first makes it impossible to prove improvement later.
Fragmented pilots, or what we could call “AI theater,” dominate many enterprises’ portfolios, producing usage data but not business data. And KPIs are frequently misaligned across the teams involved. For example, IT measures adoption, the business unit measures satisfaction, and finance measures neither, meaning that internally nobody is speaking the same language, which can lead to inaccurate ROI measurements.
The Biggest Mistake: Confusing Adoption with Value
I believe the most common mistake in measuring enterprise AI is using adoption metrics like active users, feature adoption, or query volume as a stand-in for value. These numbers are easy to collect and report, but they don’t show real business impact.
This is because these metrics only measure activity. Finance teams and boards, who answer to shareholders, care about measurable outcomes like higher revenue or better margins.
Put simply, every AI project should be judged by the business outcomes it aims to improve. If AI doesn’t move key business metrics, it will not scale.
To put this into practice, make business outcomes the first point in any AI investment proposal.
A Framework for Measuring AI ROI
The following framework should not be treated as a complete, finalized blueprint for measuring all ROI from AI initiatives. But I believe it’s a coherent starting point that can be iterated upon as required to suit most AI projects.
1. Start With Material Business Outcomes
A credible AI measurement framework starts with a clear link between the project and a real business outcome. This could be revenue growth, cost reduction, or customer lifetime value.
Before spending any money, every AI project should be able to fill in this sentence: “This program will move [specific metric] by [target amount] within [timeframe], and here is how we will measure it.”
What this discipline eliminates is innovation for innovation’s sake. As a business investment, AI deserves the same rigor applied to any business investment decision.
2. Establish a Baseline Before Deployment
ROI without a baseline is just a story. Before launching any AI project, document the current state of the process you want to improve. Use clear metrics like sales cycle time, cost per transaction, or customer churn rate in the target area.
Often, in the rush to deploy new tools, this step is skipped. When results come in, but the team can’t show what changed because there was no baseline, the real value created is hard to prove. Setting a baseline builds credibility for every future discussion about AI value.
3. Measure Workflow-Level Impact
If an AI tool speeds up a single task, that is helpful but limited. The bigger impact comes at the workflow level: did cycle time improve from start to finish, did decisions get faster, and did throughput increase across all steps?
This matters for measurement because value compounds when entire workflows improve. A 15% improvement in each of five steps within a workflow produces a system-level improvement far greater than the sum of its parts. Measuring only at the task level misses the most important part of the story.
4. Quantify Both Direct and Indirect Value
Direct value is easier to measure. It shows up as increased revenue from AI-assisted sales or cost savings from automation. These numbers are straightforward to present in a financial review.
However, with AI, indirect value often creates long-term competitive advantage. This value is harder to measure, but it should not be ignored. Over time, indirect value can make up a larger part of AI’s total return than direct savings. To measure things like improved user satisfaction from better onboarding or faster error resolution, teams need to create clear metrics and work closely with those who know the business best.
5. Track Compounding Value Over Time
Perhaps the biggest change in thinking is moving from a one-time ROI snapshot to tracking ROI over time. Traditional business investments deliver returns and then end. AI is different: it gets better with more data, feedback, and scale. An AI system might improve a workflow by 10% in the first quarter, but by 18% at the end of the year as it learns and integrates more deeply.
Measuring AI as a one-time gain systematically understates its value and leads to underinvestment precisely when the compounding returns are beginning to accelerate. The measurement framework must capture value over time and what the trajectory suggests about the long-term value of continued investment.
Communicating AI Value to the Board
The measurement framework is only as valuable as the organization’s ability to communicate what it reveals. For CIOs, this means a fundamental shift in how AI outcomes are narrated into business impact language.
It’s not 2001, and boards today aren’t unsophisticated about technology. What they’re rightly skeptical of is technology investments that can’t articulate their contribution to financial performance with the same clarity expected of any other strategic business investment, as they’ve been burned by high-profile cases where this exact question went unasked or unanswered, leading to enormous shareholder value destruction. The CIO who can speak that language fluently is the one who earns the trust and the budget to scale.
Funding discipline is just as important as measurement discipline. The best approach is a ‘pilot, prove, scale’ model:
- Defined investment at each stage,
- clear milestones that must be met before the next tranche is allocated,
- and a commitment to reinvesting in the next phase of AI deployment.
Some may resist short feedback cycles and clear milestones, but these features make ongoing, large-scale AI investment possible over many budget cycles. Treat AI like any other strategic investment, and it will receive similar funding.
The CIO’s Role in Driving Measurable Value
The evolution of the CIO role from cost center manager to value creator is well-documented in theory. But unlike more settled roles within the business’s leadership structures, it still seems to be inconsistently executed in practice.
This inconsistency affects whether AI adoption succeeds or fails. Right now, budgets are open to new AI spending, but this won’t last. When budgets tighten, only AI programs that show real business value will survive. CIOs who cannot prove this value will not lead the next wave of enterprise change.
You can recognize a strong AI program by a few signs: AI is built into core workflows, value is measured consistently over time for both direct and indirect returns, and adoption grows because of real outcome improvements, not just new features.
This is what repeatable value creation looks like at the enterprise level. And it’s the standard against which AI programs will increasingly be evaluated. The CIO as “Chief Impact Officer” is probably a better description of what the role must now actually deliver.
From Experimentation to Enterprise Value
Showing the shift from AI experiments to real business value using business metrics and outcomes is the main challenge for tech leaders today. The real work is making this part of daily operations and moving from “This works in theory” to ‘This is how we drive better results.”
The markers of that transition are specific and identifiable:
- AI is embedded in core workflows (not just reporting dashboards).
- AI is structurally present at the decision points and handoffs that drive business outcomes.
- There are consistent value measurement metrics present that capture workflow-level and compounding returns, not just early-cycle task improvements.
- Scaled adoption that is tied to outcome achievement, so that expansion of the program is earned through demonstrated impact rather than assumed from technical capability.
Enterprise AI success may come with less excitement than we have seen in recent years. I believe true success looks like repeatable value creation and the ability to use AI in important workflows, measure its results, learn from them, and apply those lessons to future projects.
Once built, this ability is one of the most lasting competitive advantages a company can have.
Key Takeaways
I believe the organizations that see this period as a turning point will not be those that ran the most pilots or used the most tools. Instead, they’ll be the ones who treated AI like any important business investment by setting clear goals, strong baselines, outcome-based measurement, and real accountability.
True AI success will be an impact story. The technology is clearly compelling, but compelling technology that can’t demonstrate measurable returns is just another cost, not an asset.
This means expectations are changing. Boards and CFOs no longer accept activity metrics or innovation stories. They want the same rigor from AI as from any other investment, and this is what will allow for ongoing, large-scale AI investment.
The real question is not whether you use AI, but whether it delivers measurable, growing value on your business investment, and whether you can explain that clearly and simply.
