Your Freed Hours Are Not Money

A CFO once ended a business case I was sitting in on with one question. The deck said the automation would free 900 hours a year, and at a loaded rate of $95 an hour, that was $85,500 in annual savings against a build that cost a third of it. Everyone liked the slide.

She asked which line of the P&L the $85,500 was coming out of.

Nobody could answer. There was no headcount reduction planned. Nobody was going to work fewer hours. The 900 hours were real, in the sense that the work would genuinely stop happening, but the money was a translation somebody had performed on a spreadsheet, and it did not correspond to a dollar anyone was ever going to see.

She was right, and the slide was wrong in a way that almost every AI business case is wrong.

Time Saved Is Not the Same Thing as Money Saved

I argued in Why Most Companies Are Getting AI ROI Wrong that hours saved is a vanity metric and decisions improved is the real one. I want to extend that, because there is a specific arithmetic error underneath it that keeps producing business cases that fall apart the first time a finance person reads them carefully.

The error is treating an hour of freed capacity as equivalent to an hour of cash. They are not equivalent, and there are three reasons why.

A large share of the time never gets freed at all. Workday's January 2026 research, covering 3,200 employees at companies above $100M in revenue, found that 85% report saving one to seven hours weekly, yet nearly 40% of those savings are lost to rework, meaning correcting errors, rewriting output and verifying results. The same study found 77% review AI-generated work just as carefully as human work, if not more, and that only 14% consistently achieve clear, positive net outcomes. Your 900 hours were gross. The net is smaller and the difference is not a rounding error.

What is left over usually gets absorbed rather than banked. Upwork's research found that 77% of employees say AI tools have added to their workload, and that 47% have no idea how to achieve the productivity gains their employers expect of them. That is from 2024 and the specific numbers should be read with their date attached, but the shape has held. Freed time in a busy department does not sit in a jar. It flows into the next thing on the list, which may be more valuable than what it replaced, or may not.

Nobody counts the new work you just created. Someone has to own the automation, watch it, and fix it when the source system changes its export format in March. That is real recurring effort and it almost never appears in the case.

What an Honest Number Looks Like

None of this means the value is fake. It means the translation from hours to dollars needs a step that most cases skip.

The step is a capacity valuation rate: what fraction of freed hours actually converts into something the business can point at. Set it explicitly, argue about it out loud, and default it low. In the models I build, the default is 50%, and I show the case at 25% and at 100% alongside it so the reader can see how much of the conclusion is riding on that one assumption. If a case only works at 100%, it does not work.

Then be specific about what the recovered capacity turns into, because there are only a few honest answers. Sometimes it is headcount you do not have to add, which is a real number a CFO recognizes. Sometimes it is throughput, more bids submitted or more accounts touched, which is real but belongs on the revenue side with a conversion assumption attached. Sometimes it is cycle time, which shows up as a better win rate rather than a lower cost. And sometimes the honest answer is that it turns into relief, which is worth having and is not a financial return, and saying so is better than dressing it up as one.

Finally, put your own fee inside the first year. A build that costs $30,000 against $85,500 in gross annual savings looks obvious. The same build against $42,000 of realistically valued savings, minus the fee, minus a maintenance allowance, is close to break even in year one and clearly positive in year two. That is a real business case. The first version is a sales document.

Why the Dishonest Version Costs More Than It Wins

You can get a project approved on the inflated number. You will get exactly one.

McKinsey's 2026 State of AI survey found only 37% of organizations attribute any EBIT impact at all to AI, unchanged year over year, with just 6% seeing material impact. There are several reasons for that gap, but one of them is uncomfortable and rarely said: a lot of projects that were approved on hours-saved arithmetic worked exactly as designed and still produced no financial result, because the arithmetic never described a financial result in the first place.

That failure is worse than a project that does not ship. A project that does not ship costs you the project. A project that ships, works, and cannot be shown to have mattered costs you the credibility of the next four proposals, and it teaches your finance team that AI cases are marketing documents.

The teams I see doing this well are the ones whose first business case was less impressive than it could have been. They discounted the capacity, they named what the freed hours would actually become, they put the fee inside year one, and when the number came out marginal they said so and picked a different process. The second and third cases got approved faster because of it.

Try This on the Case Sitting on Your Desk

Take whatever AI business case you are currently holding and do four things to it.

Cut the hours-saved figure by the share that will be spent reviewing and correcting the output, and be generous about that share. Multiply what remains by an explicit capacity valuation rate, and write the rate on the slide where people can argue with it. Name what the recovered capacity becomes in words a CFO would recognize, and if you cannot, say that too. Then add the annual cost of somebody owning the thing.

If the case still works, you have something you can defend in a room. If it does not, you have saved yourself a project that was going to be measured against a promise it could never keep.

The Takeaway

Freed hours are the beginning of a business case, not the end of one. Somewhere between the hours you save and the money you keep sits rework, absorption, and the cost of owning what you built, and every one of those is a real subtraction.

Do the subtraction before your CFO does. Discount the capacity on purpose, say what the time actually becomes, and put your own cost inside the first year.

A smaller number you can defend is worth more than a bigger one you cannot, because the smaller one is the only kind that survives the meeting where somebody asks which line it comes out of.

Tracy Thayne* is the founder of Expona, an AI-powered operational intelligence platform for B2B marketing. Read the Expona founder story or subscribe to the blog (below) for weekly insights on context, AI, and the operating model of the next decade.*