AI ROI Measurement - Half of Businesses Can’t Do It
AI ROI measurement is the challenge that’s quietly undermining confidence in artificial intelligence across boardrooms globally.
A January 2026 IDC study of 1,317 senior AI decision-makers found that 50.8% of organisations want to measure the return on their AI investments but can’t do so due to limitations in their current setup. Nearly a third — 29.8% — predict this will be their biggest challenge over the next two years.
AI is delivering results for businesses that deploy it properly but if you can’t prove it, you can’t justify it, fund it or scale it.
Why AI ROI Measurement Is So Difficult
The problem isn’t that AI doesn’t work. The IDC research shows it clearly does. Among surveyed organisations, 28.9% reported a 14.4% average improvement in workforce productivity. Some 31.8% saw operational processes speed up by 10.1%. And 28.3% reported a 14.1% improvement in customer satisfaction.
Those are significant, commercially meaningful numbers. So why can’t businesses prove it?
The IDC study identifies a fundamental disconnect between the two types of metrics organisations need to track. Business metrics focus on strategic impact: revenue growth, cost reduction, operational speed. Technical metrics focus on model performance: accuracy, latency, throughput. Most organisations struggle to connect the two for AI ROI measurement.
Data scientists can tell you a model’s accuracy improved by 3%. What they often can’t tell you is whether that translated into £50,000 in saved costs or 200 fewer customer complaints. The business case for AI lives in the translation between these two worlds and most organisations haven’t built the framework to make that translation happen.
The Metrics Businesses Want But Can’t Track
The IDC study asked organisations what they’d like to measure but currently can’t and the results reveal exactly where the gap sits.
The top metric organisations want is “intelligence per dollar” — the total cost per unit of useful AI output delivered to an end user. Some 53.9% of respondents want this but can’t measure it. It’s a powerful concept: not just how much AI costs, but how much useful work it actually produces per pound spent.
ROI came second at 50.8%. Then cost per token at 40.7%, which matters increasingly as businesses scale generative AI usage and every query has a direct infrastructure cost.
Further down the list: time to insight (40.1%), reduction in operational costs (36.1%), speed of process (24.1%), employee hours saved (17.5%) and percentage of GPU utilisation (15.3%).
Businesses are investing in AI but flying blind on whether that investment is delivering proportionate value. They know they’re spending and they believe they’re benefiting but they can’t quantify it with enough rigour to defend the budget at board level.
Why AI ROI Measurement Matters More Than Ever
The IDC study found that controlling the rising costs of AI is the top concern for 32.6% of organisations over the next two years. When budgets are under pressure — and they are — every line item gets scrutinised. AI investments that can’t demonstrate clear returns are the first to get cut.
This creates a dangerous cycle. Businesses invest in AI, see qualitative improvements but can’t quantify them. When budget reviews come around, the AI programme can’t defend itself with hard numbers. Funding gets reduced or reallocated. The AI initiative stalls and the business falls behind competitors that maintained their investment.
The organisations that avoid this trap are the ones that build measurement into their AI strategy from the very beginning and not as an afterthought once the tools are deployed.
How to Build AI ROI Measurement Into Your Strategy
The IDC whitepaper recommends bridging what it calls the “boardroom gap” — the disconnect between technical AI performance and business-level value. This requires three things.
First, defining success in business terms before you deploy. If you can’t articulate what AI should achieve in language your board understands — cost savings, revenue uplift, hours saved, customer retention — then you can’t measure it later. The metrics need to be agreed upfront, not reverse-engineered after the fact.
Second, implementing telemetry that tracks both technical and business KPIs. This means monitoring not just model accuracy and latency but the downstream business outcomes those metrics drive. How many support tickets were resolved? How much processing time was saved? What was the cost per inference?
Third, using real-time monitoring and automated reporting so the data is always current and always available to decision-makers. Annual AI reviews aren’t sufficient. Leadership needs ongoing visibility into what AI is costing and what it’s delivering.
This is exactly the kind of strategic groundwork that an AI Workshop establishes. It defines what AI success looks like for your specific business, in measurable terms, before any technology decisions are made. An AI Roadmap then builds measurement into the implementation plan, ensuring that when AI is deployed the ROI case is trackable from day one.
The Bottom Line
AI ROI measurement isn’t a nice-to-have. It’s the difference between an AI programme that survives its second budget review and one that doesn’t.
The IDC data is stark – over half of businesses investing in AI can’t measure what they’re getting back. In a climate where every cost is scrutinised, that’s unsustainable. The businesses that will scale AI successfully are the ones that define success upfront, measure relentlessly and can walk into a board meeting with clear, quantified evidence of value.
If you can’t do that today, you need to fix it before your next investment round, not after.
Complete our free AI Readiness Assessment to start building the measurable foundations your AI strategy needs.
