From Productivity To Capability: Rethinking AI Value
Luboslava Uram is COO and CTO at Solvd Group, a subsidiary of Allianz Group. Transforming the claim management experience.
gettySooner or later, almost every discussion about AI turns to productivity. How many hours did we save? How much faster are people? How much cost can we remove?
These are fair questions, and as COO and CTO, I ask them myself. We need to understand whether technology really creates value and whether the investment makes sense. But I increasingly think productivity is only the first layer of the AI discussion.
There is enough evidence that AI can make people faster. A large field study involving more than 5,000 customer-support agents found that access to generative AI increased productivity by around 14% on average. So, the productivity case is real. What I question is whether this is where the biggest value will come from.
We have seen many technologies improve productivity before. Email made communication faster, ERP helped us standardize processes and information, workflow automation reduced manual activities. AI does all of this as well, but it can also change the amount and complexity of work an organization is able to handle.
The World Economic Forum found that 86% of employers expect AI and information-processing technologies to transform their business by 2030. Microsoft’s 2026 Work Trend Index gives another interesting signal: 66% of surveyed AI users said AI allows them to spend more time on higher-value work, while 58% said they are producing work they could not have produced one year earlier.
For me, this is the important difference. Productivity asks whether we can do today’s work faster. Capability asks whether we can now do something that was not realistically possible, economical or scalable before.
This becomes very visible in operations. Imagine an organization processes one million transactions. The normal AI business case asks whether we can reduce processing time per transaction. Of course we should ask this, but we should not stop there.
A much more interesting question is whether AI allows us to analyze every transaction instead of only a sample, continuously identify unusual patterns, combine information from several systems, or escalate only those cases where human judgment is really needed.
The problem with measuring hours saved is that it’s too easy. We automate one step, save five minutes, multiply it by transaction volume and employee cost, and we have an ROI. It is clean, understandable and easy to present. But we can create hundreds of such use cases and still operate almost exactly like before.
McKinsey’s recent research points in a similar direction. AI adoption is already broad, but many organizations are still working with isolated pilots rather than seeing material enterprise impact. One of the more useful recommendations from this research is to stop looking only at hours saved and ask instead what the business is now willing and able to attempt.
I think this is the right direction.
I would not remove productivity metrics. I would simply add another dimension. For every important AI initiative, I would still want to know what happens to time, cost per transaction, throughput and quality.
But I would then ask a second set of questions. How much more of our business can we now cover? Can we move from analyzing 5% of transactions to 100%? Can we handle cases that were previously too complex for automation? Can our experts supervise a much larger volume because routine work is handled differently?
These are capability questions, but they are not soft questions. They can also be measured.
Instead of saying, “AI improves decision making,” I would want to see what exactly improved. Instead of saying, “We have better coverage,” I would want to know whether we moved from 10% to 50% or to 100%. Instead of saying, “AI supports experts,” I would want to understand how many cases still require expert intervention, and whether the quality stayed stable or improved.
We need to become as disciplined in measuring new capability as we already are in measuring cost. Otherwise, capability becomes just another nice word in an AI presentation.
There is one question I would add to almost every AI investment discussion: What constraint disappears if we implement this?
This is, for me, much more useful than asking only how many hours we save. Once we understand which constraint AI removes, we start seeing what the business can do differently afterwards. This is where AI becomes strategic rather than only productive.
I am not arguing against productivity. Quite the opposite. As COO and CTO, I care a lot about productivity, quality, capacity and cost. But I think productivity should be the beginning of the AI business case, not the end of it.
Stanford’s 2026 AI Index continues to show fast progress across reasoning, coding, multimodal and agentic capabilities. As these capabilities become stronger, the value of AI is more likely to move from helping one person perform one task faster toward changing what the whole organization is able to do.
So, perhaps boards need to look at AI through two lenses. The first is efficiency: Are we doing today’s work better? The second is capability: What can we now do that we could not do before?
For me, the second question will become more and more important. It might be asked by the companies winning with AI.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
