Larridin Blog

Why Most AI Measurement Is Measuring the Wrong Thing | Larridin

Written by Larridin | Jul 12, 2026

79% of executives see productivity gains from AI. Only 29% can confidently measure AI ROI. That gap is a measurement design problem, not a data problem.

Key Takeaways

  • IBM's Think Circle report found only 29% of executives say they can measure AI ROI confidently, while 79% see productivity gains, meaning value exists but the measurement to prove it financially is broken.
  • Only 25% of AI initiatives have delivered expected ROI, and 16% have scaled enterprise-wide, according to IBM’s CEO study.
  • Organizations that measure AI after implementation and report results are far more likely to translate use cases into consistent enterprise value.

The 3 Things Most Organizations Measure Instead of Impact

The measurement instinct for AI is understandable. You deploy a tool and want to know if it’s working. So you measure what’s easiest to see: licenses deployed, users logged in, prompts processed. Those are activity metrics. They tell you AI is being used. They don’t tell you whether that usage is generating business value.

 

The distinction matters enormously when you are trying to build the financial case for continued investment. Our AI measurement framework guide breaks down the full framework for closing this gap. Here is why the most common measurement approaches fall short:

 

1. License deployment rates

Buying 500 Copilot seats and reporting that 500 employees have access to AI is just procurement confirmation. It doesn’t measure ROI. Whether any of those employees are using the tool effectively to produce measurable business outcomes is a completely different question that license counts can’t answer.

2. Login frequency

An employee who logs into an AI tool twice a day for five minutes is counted as an active user with the same weight as one who has integrated AI into every significant workflow they manage. Login frequency conflates presence with productivity.

3. Self-reported time savings

Survey-based productivity estimates are systematically biased upward because employees who are enthusiastic about AI overestimate their time savings and employees who are skeptical underestimate them. The AI proficiency measurement approach captures behavioral signals from actual tool usage rather than asking employees to estimate their own productivity improvement.

What Impact Measurement Actually Requires

Measuring AI impact rather than AI activity requires a three-dimension AI measurement framework that covers all of Utilization, Proficiency and Value together, not any one of them in isolation.

1. Utilization: who uses AI, how often and in what workflows

The AI Adoption dashboard captures this at the team and role level in real time, showing not just aggregate adoption numbers but the specific workflows where AI is integrated versus where it sits unused despite access.

2. Proficiency: how well is AI being used

Utilization without proficiency produces activity, not value. The AI Fluency measurement captures proficiency through behavioral signals, prompt sophistication, feature adoption depth and workflow integration, without requiring content monitoring.

3. Value: what business outcomes has AI changed

This is the layer that converts AI measurement from a reporting exercise into a strategic tool. The AI Impact platform connects usage and proficiency data to business outcomes: cycle time, throughput, output quality, revenue influenced and cost per outcome. That connection is what makes AI ROI calculable rather than estimable.

Why This Matters at the Pilot Stage

Only 25% of AI initiatives have delivered expected ROI, and 16% have scaled enterprise-wide, according to IBM’s CEO study. Pilots that don’t establish measurement baselines at launch have no data to build the business case for scaling. When the scale decision arrives, the team is left arguing from anecdote and adoption metrics, which is exactly the kind of evidence that gets pilots cut in favor of alternatives with clearer financial evidence. The board-ready AI ROI approach requires measurement to start at the pilot, not after the production launch.

Frequently Asked Questions

What’s the difference between AI activity metrics and AI impact metrics?

Activity metrics measure what AI’s doing: licenses deployed, users logged in, queries processed, time spent in tools. Impact metrics measure what AI is changing: business outcomes, productivity shifts, cost reductions, revenue influenced. Activity metrics tell you AI is being used. Impact metrics tell you whether that usage is creating value.

Why do most organizations struggle to measure AI ROI?

Because they measure inputs and activity instead of outcomes. IBM’s Think Circle data shows the problem: 79% of executives see AI productivity gains, but only 29% can measure ROI confidently. Leaders can see AI is helping, but they still can’t prove the return. Closing that gap requires measurement infrastructure that connects usage to business outcomes.

What is the Utilization, Proficiency and Value framework?

It’s the three-layer measurement architecture that Larridin uses to connect AI activity to business impact. Utilization captures who uses AI, how often and in which workflows. Proficiency captures how effectively they use it. Value captures the business outcomes that result. Together they produce the complete picture that boards and CFOs require. The full framework is covered in our AI measurement frameworks guide.

When should measurement infrastructure be put in place?

At the pilot stage, not after the production launch. Organizations that establish measurement baselines before AI deployment have the before-and-after data required to build a financial ROI case when the scale decision arrives. Organizations that add measurement after the fact are always working from incomplete evidence.

Measure What Actually Moves the Business

Larridin's three-dimension framework, Utilization, Proficiency and Value, is built specifically to close the gap between activity metrics and the business outcomes that boards and CFOs actually need to see.

 

Book a discovery call to see how impact measurement works.

  • AI Measurement Frameworks Guide
  • Efficiency Gains Aren’t Enough: The Shift to P&L-Level AI Value
  • Building Continuous Measurement Into AI Strategy
  • The Larridin Guide to ROI for Enterprise AI