Skip to content

The Team Says AI Saved 40 Hours. How Do I Check?

November 12, 2026

Someone on your team says AI saved 40 hours this month. It might be true. It might also be a guess that sounded good in a meeting.

You can find out in about an hour. You don’t need special software. You need four questions and a stopwatch.

This post is about checking a claim someone already made. If you want to test a task before you set anything up, read You Tried AI and It Became Another Job first.

Why “40 hours” is usually the wrong number

Most time-saved claims come from one subtraction. The old way took this long. The AI way takes that long. The difference is the saving.

The trouble is what goes into the second number. People count the seconds the AI takes to produce something. They forget the time spent reading it, fixing it, and cleaning up when it got something wrong. Nobody lied. They just stopped the clock too early.

A claim can also be off in the other direction. The old-way number is often a guess too. “That used to take us forever” is not a measurement.

Four questions to ask

Ask the person who made the claim. Ask kindly. You are not trying to catch anyone. You want to know what the number is made of.

Question 1: How long did the task take before, and how do you know?

You want a timed sample, or at least a count of tasks multiplied by a rough time each. “About 20 minutes per quote, and we send around 120 a month” is a real answer. “It felt like a lot” is not.

Question 2: What does the AI time include?

Does it include reading the output? Editing it? Sending it? If the answer is “the AI writes it in a minute,” the number leaves out the human part.

Question 3: How often does the AI get it wrong, and what happens then?

Every tool makes mistakes. What matters is how many reach a customer and how long they take to fix. One wrong price on a quote can cost more time than ten right ones saved.

Question 4: Is the amount of work the same?

If the team sent 120 quotes before and 200 now, some of the “saved” time went into extra volume. That can be good news. It is a different claim.

A worked example

This is a made-up example to show the arithmetic. It does not describe a real business.

A small contractor’s office says AI saved 40 hours a month on quote emails. We ask the four questions.

  • Before: 120 quotes a month at about 20 minutes each. That is 40 hours. So the 40 in the claim is the whole old workload, which means it assumes the new way costs nothing.
  • After: someone read and edited each AI draft, about 6 minutes per quote.
  • Errors: 9 quotes a month went out with a wrong detail, such as an old price. Fixing each one, including the follow-up with the customer, took about 30 minutes.
  • Upkeep: about 2 hours a month updating the price list the AI works from.
Line Hours per month
Old way: 120 quotes at 20 minutes 40
Reading and editing AI drafts: 120 at 6 minutes 12
Fixing mistakes: 9 at 30 minutes 4.5
Keeping the price list current 2
New way, total 18.5
Hours saved: 40 minus 18.5 21.5

The AI still saves time. It saves about 21 hours, not 40. That is a good result you can defend, and it tells you where to improve. Fewer errors would give back up to 4.5 hours.

How to run your own check

You can do this on any task your team says AI helped with.

  1. Pick one task. Do not audit everything at once.
  2. Write down the claim in hours per month.
  3. Count how many times the task happened last month.
  4. Time five of them yourself or ask someone to. Use a phone stopwatch. Include reading, editing, and sending.
  5. Count the mistakes you know about. Ask the team how long each one took to fix.
  6. Add any regular upkeep, like updating instructions or reviewing what the tool did.
  7. Fill in a table like the one above with your own numbers.

If your total lands near the claim, good. If it lands at half, you still have a tool that works. You just have a more accurate picture of it.

Signs the number is soft

  • Nobody can say how many times the task happened.
  • The old-way time is a feeling, not a count.
  • Nobody has checked the AI’s output in a month.
  • The savings are described in total hours for the whole team, with no task named.
  • The person claiming the saving is also the person who set up the tool, and nobody else has looked.

None of these mean the claim is false. They mean you can’t tell yet.

What to do with the answer

A confirmed saving is worth keeping and worth writing down, so next quarter you are comparing to a real baseline.

A smaller saving still counts. Look at the biggest line in the table. Usually it is the review time or the mistakes. Both can often be cut by giving the tool better instructions or narrowing what it is allowed to do.

A saving that disappears is useful to know too. Switching a tool off after you counted honestly is a sound decision.

How we would approach it

If you asked us to check a claim like this, we would start with the same four questions, then time a sample of the real work and build the table with your numbers. We have no client results to point to here. The method is the same arithmetic you can do yourself, and you are welcome to do it without us.

Keep reading

Want to talk it through?

Tell me about the task that takes up your team's time.

Newsletter

Plain-language notes on software and AI for small business

One short email when there is something worth sharing. No spam. Unsubscribe any time.