One team says it has cut the time it takes to prepare reports in half. Another finishes its presentations faster. Customer replies are drafted in a few minutes. These are welcome developments. But in the management meeting we need to ask a somewhat harder question: which of the company’s results did these gains change? Did the customer have a better experience, did the backlog shrink, did quality improve, did a new revenue opportunity emerge?
When we think about the economic return on AI investments, time savings are a strong starting point. But the link between time and economic value does not build itself. Finishing a task faster can free up capacity. Where that capacity is used, the other bottlenecks in the workflow and whether the resulting output finds demand are separate management decisions. The area I care about is this gap between acceleration and realized value.
Gross time, net capacity
Let us think through an entirely hypothetical example. A team of twenty reports saving thirty minutes per person per working day. Over a month of twenty working days, that amounts to two hundred hours of gross capacity. If checking and correcting the AI output takes an extra ten minutes per person per day, net capacity falls to roughly one hundred and thirty-three hours. I am not presenting this calculation as a real company result. My aim is to show how the scope of measurement changes the result within the same example.
Scattered minutes and bottlenecks
Now to the second question. Are these one hundred and thirty-three hours actually usable? Small fragments of time scattered across the day do not always add up to uninterrupted capacity that can take on new work. If someone saves four minutes at five different points, they may not be able to use that time for a long customer analysis. If the team can reorganize its work and pool this time, the resulting capacity becomes more meaningful. This is exactly where work design becomes an economic issue.
The third question is where the capacity flows. If you speed up proposal drafts but price approval still waits several days, the total time the customer experiences may hardly change. A department that works faster can also build a longer queue in front of the next bottleneck. That is why looking only at task time falls short. We need to see the end-to-end time of the work together with the result that reaches the customer.
A customer support study reported by MIT Sloan in 2024 offers a useful example for thinking about this distinction. In the study, access to an AI tool is associated with an average productivity increase of about 14 percent, measured as issues resolved per hour. It matters that the metric is tied directly to work output. Still, we cannot conclude from this finding in a single work setting that every company’s profitability will rise by the same rate.
Different paths to value
Managers have different paths to value. Freed capacity can make it possible to meet more demand. It can be devoted to higher-quality contact with existing customers. It can be used to find errors earlier, solve recurring problems or make time for employees to learn. Each path has a different outcome and a different measure. Being able to do more work produces an economic return only if demand and other resources allow it. Time devoted to quality, on the other hand, shows its effect in a different indicator.
That is why I am cautious about automatically multiplying the time saved by personnel cost and writing the result into the profit column. Such a calculation can show the approximate cost equivalent of the freed capacity. A claim of cash savings requires spending to actually go down. Likewise, additional sales potential is not revenue until the sale happens. Keeping the distinctions between capacity, cost, revenue and customer value does not belittle the AI investment; it makes the assessment more reliable.
The agreement with employees
The agreement with employees is also decisive here. If we want the team to make the time it saves visible, we need to talk about how that time is used. A manager might, for example, set aside part of the week for examining recurring customer complaints. The team channels the time freed up by AI into this work. Then we look at whether complaint recurrence, the need for rework or resolution time has changed. In this way, a traceable link is built between acceleration and outcome.
This approach has an important limit. Not every benefit can be turned into money quickly or easily. A lighter cognitive load for employees, better-prepared decisions or a more consistent service language can be valuable. It would also be wrong to ignore these effects just because they do not immediately fit into a financial measure. What is needed is to track each effect with its own appropriate indicator and to say openly that the evidence required for a monetary claim does not yet exist.
A simple measurement routine for the pilot
In a pilot, a simple but disciplined measurement routine may be enough. We select comparable tasks and record the times before and after AI, including the effort spent on checking. We review quality against the same criteria. Then we track which work the freed capacity went to and what that work achieved. If the difficulty of the tasks has changed, if another process improvement happened in the same period or if the sample is small, we state these limits alongside our interpretation. We do not label every positive difference an AI effect.
A new column: the capacity-use decision
My suggestion is to add a single new column to the AI review meeting: “Decision on how the freed capacity will be used.” If this column does not contain a task, an owner and an observable outcome, what we have is a value story that is not yet complete. The tool may have accelerated part of the work; leadership’s job is to decide which priority this speed will serve.
At your next investment review, do not ask only how many hours were saved. Ask also which customer need, quality problem or growth opportunity those hours were devoted to. Because what time contributes to an organization depends as much on what we put into the space that opens up in the calendar as on the space itself. Is your company consciously managing how it uses the time it gains today?
Sources
- MIT Sloan, Generative AI and Worker Productivity (16 April 2024)
- Erik Brynjolfsson, Danielle Li, Lindsey Raymond, Generative AI at Work, NBER Working Paper 31161