Companies globally are pouring trillions into artificial intelligence, but a surprising number struggle to see a real return. The problem often lies not with the technology, but in how its success is measured. It’s time to look past usage and focus on results.
The Flaw in Counting Clicks
For many organizations, the primary metric for AI success has been adoption. Dashboards proudly display how many employees have logged in, how many queries they've run, or how many teams are 'using' AI tools. While these figures show engagement, they reveal nothing about value. This focus on activity creates what McKinsey calls the 'Generative AI Value Paradox': while around 80% of companies report using AI, a similar percentage admit they haven't seen a significant impact on revenue or costs. Simply put, being busy with AI is not the same as creating business value with AI. This gap highlights the urgent need for a more sophisticated way to measure return on investment, moving from 'Are we using it?' to 'What are we achieving with it?'.
A Framework for Real Value
To bridge this gap, McKinsey proposes a structured approach to connect AI deployment to financial results. One recent framework uses five distinct layers of measurement, creating a clear line from technical performance to bottom-line impact. The layers start with the technical foundation (Is the model working correctly?) and move up through user adoption (Are people using it in their workflows?), operational KPIs (Is work getting done faster or better?), strategic outcomes (Are we gaining market share?), and finally, financial impact (Are revenues up or costs down?). This pyramid structure ensures that investments are validated at each stage. A project only advances when it shows credible evidence of success at its current level, preventing companies from scaling expensive initiatives that only look good on a usage report.
Identifying What Truly Matters
The key to this outcome-based approach is redesigning entire processes instead of just applying AI to isolated tasks. Rather than automating one small step in a ten-step workflow, successful companies reimagine the entire workflow with AI embedded. This means organising cross-functional teams around a specific business outcome, like 'improving customer lead generation' or 'reducing product development time'. For example, one company used AI to enhance its lead generation process and doubled the average order value from its initial prospects. The focus wasn't on how many salespeople used the AI tool, but on the direct increase in order value. This requires business leaders to understand AI's potential and tech teams to understand business needs, ensuring use cases are chosen for their impact, not just their feasibility.
The Trillion-Dollar Estimate
When this outcome-focused approach is applied correctly, the potential is enormous. McKinsey's research estimates that generative AI could add the equivalent of $2.6 trillion to $4.4 trillion in value to the global economy annually. This staggering figure isn't just hype; it’s based on an analysis of 63 specific use cases across 16 business functions. Around 75% of this value is concentrated in four key areas: customer operations, marketing and sales, software engineering, and R&D. For example, the banking industry could see an additional $200 billion to $340 billion annually, while retail could gain between $400 billion and $660 billion. These numbers illustrate that the greatest benefits come from augmenting high-skilled work, not just automating low-wage tasks.
Making the Shift in Your Organisation
Transitioning from measuring activity to outcomes requires a deliberate strategy. It starts with leadership aligning on what business KPIs the AI initiative is meant to improve before a single line of code is written. Instead of tracking logins, measure changes in sales conversion rates, customer satisfaction scores, or the time it takes to bring a new product to market. It also means being realistic about costs. A recent McKinsey analysis, 'The AI Value Curve', noted that while AI could create a potential $2.52 trillion in annual value, a 30% rise in deployment costs could erase $850 billion of that potential. By 2030, an estimated $7 trillion will have been invested in AI infrastructure, and this spending needs to be justified by real profit. Success depends on focusing investment on applications with clear economic returns and redesigning work to capture them.
