There is a line in an AI business case that always makes me smile: “hours saved.” The arithmetic can be perfectly respectable. A task takes less time, a certain number of people perform it, and the multiplication produces an attractive annual figure. I would ask what changes on Monday morning. Who does something differently because those hours now exist? If the answer is that people will “focus on higher-value work,” I would keep the meeting open a little longer.
The FII Institute’s enterprise-AI report, launched on 12 October, puts workflow redesign and workforce development among the conditions for turning AI investment into lasting business value (report launch). That is the part of the AI conversation I want to spend more time on: how managers turn a technical improvement into a different working week.
An hour in a spreadsheet is easy to move
Consider a hypothetical team of twelve people. An AI tool genuinely saves each of them half an hour a day, after checking the output. On paper, that is six hours of capacity every day. It sounds like a substantial resource. In practice, those six hours belong to twelve different people with different skills, commitments and calendars. We cannot simply pour them into a vacant role. Some of the time may be useful immediately; some may remain scattered between meetings and interruptions unless somebody reorganizes the work.
That distinction should be clear to a CIO. I would separate time saved on a task, usable capacity in a team, and a business result. Each deserves its own evidence. Treating all three as interchangeable makes the investment story easier to tell and harder to defend later.
It also risks missing a perfectly legitimate benefit. A team that can absorb demand without working late may have gained something valuable, even if no budget line has moved. I would want us to describe that result honestly rather than force every improvement into a cost-reduction claim.
Someone has to decide what gets the time
“Higher-value work” becomes useful when a manager can point to something specific. Which recurring incident will we finally investigate properly? Which customer issue needs attention before renewal? Which junior colleague needs time with a more experienced person?
Those are choices. Giving the reclaimed time to one of them means protecting it from other requests. A manager who is still expected to accept every new demand has very little room to make that protection credible. I would therefore want an AI deployment to include a conversation with the person who runs the team. What work will shrink? What will take its place? What are they authorized to postpone or stop so that the new priority survives?
This connects with something I argued in the case for a strong N-1 layer: management structure determines whether leaders have room to lead. For AI, I would apply that argument one level further. The manager closest to the work needs enough authority to redesign it, and enough support to explain the trade-offs.
The old expectations need to change too
Suppose AI helps a team prepare a weekly report in twenty minutes instead of two hours. That is a useful improvement. I would still ask who reads the report, which decisions it supports, and whether its frequency or format should change. There is an obvious temptation to use the remaining time to produce more detail. A longer report, another dashboard, a daily version for someone who previously managed with a weekly one. Before agreeing, I would want to know which decision gets better.
The same discipline should apply to people’s objectives. If we ask someone to spend their reclaimed time coaching colleagues, then assess them entirely on individual output, we have given them two competing instructions. The scorecard is likely to win.
Workday’s research published in January provides a useful, earlier warning: it reported that in 89 percent of organisations, fewer than half of roles had been updated to reflect AI capabilities (Workday’s January analysis). I would treat that vendor-reported finding as a prompt to examine our own roles and expectations, rather than assume that deploying a tool has already changed the job.
There is a people conversation here that deserves plain language. If the objective is to reduce costs, say so. If the objective is to handle growth without adding the same amount of capacity, explain that. If the intention is to improve quality or make an overloaded team more sustainable, specify how we will recognize success. I would be uncomfortable asking people to reveal exactly where they can save time while giving them only vague assurances about what follows. They are being asked to help redesign their own work. They deserve a clear account of the intended outcome and of what remains undecided.
I would also leave room for the team to challenge the original calculation. An estimate made before deployment is a hypothesis. If the work turns out to be more complex, or the usable time smaller, I would rather learn that early than make people defend a number they never produced.
Follow the time until it becomes something useful
For an initial deployment, I would keep the follow-through simple. Establish what the work takes today, agree which activity will receive the capacity, and choose an observable result. Then return after a few weeks to see whether the working week actually changed.
In an IT team, that might mean using time released from routine documentation to investigate a recurring incident. The evidence would include the investigation being completed and, over an appropriate period, whether the problem recurs. Counting faster documents would tell only the first part of that story.
Finance should be involved when we intend to claim a financial benefit. The business owner should be involved when the promise concerns service, quality or speed. The team manager should be involved throughout, because someone has to make the new arrangement work on an ordinary Tuesday.
I would like every AI business case to finish the sentence: “We will use the capacity to…” A named owner, a specific piece of work and an observable result would make a convincing ending. Without them, the hours may be real, but we still have a management decision to make.
