"It saves us ten hours a week" is the most common AI success story in small business. It's also the most misleading.

This is my opinion on how to judge an AI purchase. If you want the step-by-step measurement method, read Di-Hy's guide: How to Measure AI ROI in a Contracting Business.

My position

Ten hours saved is a fact about time. Return on investment is a fact about money. They only connect if something specific happens to the time, and in most businesses nobody decides what.

So I ignore hours saved. I ask what the time was for.

The only three outcomes that count

Saved time turns into value in three ways:

  1. It gets redirected to something that earns. The office manager now makes the callbacks that used to get skipped, and some become booked jobs.
  2. It avoids a cost. You don't hire the part-time person or pay the overtime.
  3. It reduces paid hours. The work takes less paid time and you actually cut hours.

If none of those happen, you didn't save money. You got a quieter afternoon. That can be worth it. It just isn't a return, and it shouldn't sit on the same line as dollars.

What I'd ask for instead

Before I approved any AI spend I'd want three things on one page:

  • The baseline. What the number is today, where it comes from and who pulls it.
  • The full cost. Software, setup time, the weekly review and the cleanup when it sends something wrong. Not just the subscription.
  • The decision rule. Written before launch: keep it if the number moved, fix it if something is leaking, stop it if nothing moved after a fair trial.

That last one matters most. Without a rule decided in advance, a tool nobody can justify stays on the credit card for years.

Where the trade-off is

This discipline costs a little speed. It's easier to buy the tool, feel productive and never check. I'd still make the owner write the page, because the version that needs a consultant to interpret never gets used, and the version nobody checks never gets cancelled.

I would also be honest about soft wins. Fewer errors, a less miserable role, work that stops piling up on one person: those are legitimate reasons to adopt something. Label them as that. Don't dress them up as a return you've proven.

Be wary of the vendor dashboard

Vendors report what they want you to see. Check any claimed benefit against your own CRM, bank or books. If you can't, treat the claim as a sales pitch.

If you'd like someone to set up the baseline and the scoreboard with you, Di-Hy's free audit is where I'd start.

FAQ

Do hours saved count as ROI?

Only if the time is redirected to work that earns money, avoids a cost or reduces paid hours. Otherwise it's a soft benefit.

What should I measure for an AI project?

Two or three numbers tied to money, such as response time, booked jobs or paid hours on a task, with today's value written down before launch and a review date set.


Sources: No outside statistics are cited here. This is opinion. For the full method, see Di-Hy's AI ROI guide.