# What Should Your Business Do With the Time AI Saves?

Canonical URL: https://business.theaieducator.io/posts/what-to-do-with-time-ai-saves
Publication: Dan Fitzpatrick Insights
Author: Dan Fitzpatrick
Topic: AI and the Future of Work
Published: 2026-10-31T16:40:00.000Z
Modified: 2026-10-31T16:40:51.282Z

AI saves your people time, but saved time is not money in a drawer. It slips back into the same work unless you choose where it goes. The Dividend Decision gives you four destinations and a way to pick one per job.

## In brief

Choose where the time AI saves will go before it is saved, or it gets absorbed by more of the same work. Pick one of four destinations per job: Grow, Fix, Build or Return, then name an owner and a review date four weeks out.

## Key takeaways

- A UK government trial of Microsoft 365 Copilot with 20,000 civil servants reported an average of 26 minutes saved a day, but the report says it could not identify how the saved time was spent.
- In a Danish study of 11 occupations, users reported saving 2.8% of work hours, 80% of that time went to other job tasks, and only 3 to 7% of the productivity gains reached earnings.
- Workday's November 2025 survey of 3,200 AI users at large organizations found nearly 40% of time savings lost to rework and 32% saying they simply took on more workload.
- The Dividend Decision is a one-line choice made before AI saves an hour: Grow, Fix, Build or Return, one destination per job, with an owner and a review date.
- Choose in order: if paid work is waiting, Grow; if a job keeps slipping, Fix; if someone is stretched or a skill is missing, Return or Build.
- Build points the hours at framing, judgment and ownership, the skills that get scarcer as AI improves, not at more of the tasks AI is already taking over.
- Treat every saving as a hypothesis until you have checked how much came back as rework, and tell the person whose time it is that their workload will not simply rise by the saved amount.

Decide where the saved time goes before AI saves it, because an hour with no destination gets absorbed by more of the same work. Choose one of four destinations for each job: grow the work that earns, fix the job that always slips, build a skill that gets scarcer as AI improves, or return the hour to the person who found it.

A UK government trial of an AI assistant ran on 20,000 civil servants for three months. Its authors could tell you how much time people said they saved. They could not tell you what happened to it. The report is blunt: "due to experimental constraints it was not possible to identify how time saved was spent."

That gap is the subject of this article. You can hold good evidence that AI saves your people time and still reach the end of the year with nothing to show for it, because saved time is not money in a drawer. It is minutes scattered across a dozen people's days, and it goes somewhere. The only open question is whether you chose where.

## What does the evidence say happens to the time AI saves?

Three studies, three different groups of workers, one pattern: the time is real, it is modest, and it mostly slips back into the work.

The UK Government Digital Service ran a cross-government trial of Microsoft 365 Copilot from 30 September to 31 December 2024, with 20,000 licences across twelve departments and agencies. In the survey, 7,115 users reported saving an average of 26 minutes a day, which the report extends to 13 days over a full working year. The figures are self-reported, and the report says so: they "should be considered alongside additional literature on the true value of these metrics." ([GDS, published 2 June 2025](https://www.gov.uk/government/publications/microsoft-365-copilot-experiment-cross-government-findings-report/microsoft-365-copilot-experiment-cross-government-findings-report-html))

In Denmark, the economists Anders Humlum of the University of Chicago and Emilie Vestergaard of the University of Copenhagen surveyed workers in 11 occupations, with about 25,000 valid responses in each of two rounds (November to December 2023 and 2024), and matched the answers to administrative earnings records. Users reported saving 2.8% of their work hours. Of the time saved, 80% went to other job tasks and fewer than 10% went to extra breaks or leisure. Only 3 to 7% of the productivity gains reached workers' earnings, and 17% of users said chatbots had created new work for them. ([Humlum and Vestergaard, Becker Friedman Institute working paper, May 2025](https://bfi.uchicago.edu/wp-content/uploads/2025/04/BFI_WP_2025-56-3.pdf))

Workday surveyed 3,200 full-time employees who actively use AI, at organizations with $100 million or more in annual revenue across North America, Asia-Pacific, and Europe, the Middle East and Africa, in November 2025. Nearly 40% of AI time savings were lost to rework: correcting errors, rewriting content and verifying output. Instead of using the time to build skills, 32% said they simply took on more workload. Companies were more likely to put AI savings back into technology (39%) than into employee development (30%). These are large organizations, so treat the percentages as a direction, not a forecast for a 15-person firm. ([Workday, 14 January 2026](https://en-sg.newsroom.workday.com/2026-01-14-New-Workday-Research-Companies-Are-Leaving-AI-Gains-on-the-Table))

None of the three studies says what a small business should do. Here is my reading of them. Saved time has three ways out of a business: into more of the same work, into repairing the output, and into the quiet pocket of the person who found the saving. Nobody chooses any of the three. They happen because nobody chose anything else.

## What is the Dividend Decision?

The Dividend Decision is a one-line choice an owner makes before AI saves anyone an hour: where will the time go? There are four answers: Grow (more work that earns), Fix (the jobs that always get pushed back), Build (skills that get scarcer as AI improves), or Return (less strain for the people who found the saving). Pick one per job, name an owner and a review date, because an hour with no destination is absorbed by more of the same work.

I call it a dividend because that is what saved time is: the first return on the effort of learning a new way of working. A return nobody has been told how to spend is easy to waste. This is my suggested way of thinking about it, not a finding from the research above.

| Destination | Choose it when | Name this | Warning sign |
|---|---|---|---|
| Grow | Paid work is waiting: enquiries unanswered for a day, quotes late, customers you turn away | The job, and the one number it should move | The hours go into internal admin and the enquiry pile stays the same |
| Fix | A job keeps slipping: follow-ups unsent, records months behind, an onboarding pack that lives in one head | The job, and the date it should stop slipping | A to-do list that never gets shorter |
| Build | A skill is thin and will matter more as AI improves | The person, and the skill | "Development" that turns out to be more tasks (the 32% in the Workday data) |
| Return | Someone is stretched, or the saving came from their own effort | The person, and what changes in their week | Nobody tells them, so the hour is absorbed or hidden |

Choose per job, not per business. The same office can send the quote-drafting hours to Grow and the meeting-notes hours to Return.

## How do you choose a destination for each job?

Ask three questions, in this order, and stop at the first yes.

Is paid work waiting? If customers are queuing, quotes are going out late or you are turning work away, the hours go to Grow. It is the only destination that pays for itself, so it gets first call.

Which job keeps slipping? If the follow-up email is always tomorrow's job, or the practice of how you use AI lives in one person's head, the hours go to Fix. The [Small Room Test in the fewer-than-10-people guide](https://business.theaieducator.io/posts/ai-for-businesses-with-fewer-than-10-people) is a good way to find the slipping job.

Is anyone stretched, or is a skill missing? The person who found the saving gets Return if they are the one under strain. If the gap is a capability the business will need, the hours go to Build.

Take a hypothetical 15-person accountancy practice where AI now drafts client emails and meeting notes, and the office manager estimates it saves her three hours a week. Paid work is not waiting, because the practice is full, so Grow is out. The client onboarding checklist has been "nearly done" for a year, so two hours a week go to Fix, with a finish date eight weeks away. The third hour goes back to her as a protected hour on Friday afternoons that nobody books. At the four-week review the owner asks three things: did the checklist move, did the Friday hour survive, and how much of the saving was lost to checking.

## What I Tell Business Owners

I tell owners to treat saved time as the first return on AI and to spend it on purpose.

I also tell them to watch how they react to the people who report a saving. If the reward for finding a saving is a heavier workload, people stop finding savings, or stop mentioning them. The Workday figure of 32% taking on more workload is that reaction at scale, and the Danish finding that only 3 to 7% of productivity gains reached earnings suggests staff have little reason to volunteer how much time they gained. That second point is my inference, not the authors' claim.

I advise governments and system leaders, among them the UK Department for Education, KHDA in Dubai and the Ministry of Education in Kazakhstan, which is partly why the Government Digital Service admission interests me so much: a trial of 20,000 people could say how much time was saved and could not say where it went. A business of 20 people can do better, because you can simply ask.

## Why does the Scarcity Rule point to Build?

The Scarcity Rule says that when everyone has access to the same AI, a skill matters in proportion to how much scarcer it becomes as AI improves, not how useful it is today. Three skills pass that test: framing, judgment and ownership. Framing is deciding what the work is really for. Judgment is knowing whether a result is right for this customer. Ownership is being the person whose name is on the outcome.

Build means the saved hours go into practicing those three, not into more of the tasks AI is already taking over. A plain version: if AI now drafts the first version of the proposal, the hour you got back is for the conversation with the client that tells you whether the proposal is aimed at the right problem.

Build is also the destination to be careful with when the saving tempts you to skip a hire. If you are weighing that, [the Backfill Test in the replace-an-employee piece](https://business.theaieducator.io/posts/replace-an-employee-with-ai) asks the four questions to answer first, and [the entry-level jobs article](https://business.theaieducator.io/posts/ai-and-entry-level-jobs-small-business) covers what you lose when the junior work disappears.

## What if the time is not really saved?

Treat every saving as a hypothesis until you have checked it. The GDS figures are self-reported estimates picked from bands, and the Workday data says nearly 40% of the gain can leak out as rework. A saving that comes back as correction is a transfer of work from one person to another, not a saving.

Two companion pieces help here. [How to measure whether AI is working](https://business.theaieducator.io/posts/measure-whether-ai-is-working) asks you to write down, before AI goes into a job, how long it takes, how often it comes back for correction, and what the business would do with the time. This article is the next step on that third question: it gives you four places to choose from. And [training staff to catch AI mistakes](https://business.theaieducator.io/posts/train-staff-to-check-ai-output) deals with the rework leak, using a twenty-minute drill.

If you have given people access to AI and nothing has changed on the books, [why small businesses are not getting value from AI](https://business.theaieducator.io/posts/why-small-businesses-are-not-getting-value-from-ai) describes how the time leaks away before it reaches the business.

## What can you do this week?

You can run the Dividend Decision on one job in an afternoon.

1. Pick one job where AI already saves someone time. Choose the job, not the tool.
2. Ask the person for their honest estimate in minutes per week, and write it down as a guess.
3. Run the three questions above and choose one destination.
4. Write one line: "These hours go to [destination], owned by [name], reviewed on [date four weeks away]."
5. Tell the person whose time it is, and tell them what will not happen: their workload will not simply go up by the saved amount.
6. At the review, check three things: did the hours go where you said, how much came back as rework, and what did the person say.

If you already hold a weekly team conversation, [the Friday Four from the managing-a-team guide](https://business.theaieducator.io/posts/how-to-manage-a-team-that-uses-ai) has a question that fits: what should we stop doing by hand? The Dividend Decision is what you do with the answer.

## Where to go next

If you are planning a leadership offsite, a members' event or a conference session on how AI changes the way your people spend their week, that is what [my business keynotes](https://theaieducator.io/business-keynotes?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=what-to-do-with-time-ai-saves) are built for. If you would rather see where your business stands first, start with the [Business AI Readiness Scorecard](https://theaieducator.io/business-ai-readiness?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=what-to-do-with-time-ai-saves), and if your team is using AI but nobody has shown them how to use it well, [AI training for teams](https://theaieducator.io/ai-training-for-business?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=what-to-do-with-time-ai-saves) is where I would start.

## Sources and further reading

- Government Digital Service, Microsoft 365 Copilot Experiment: Cross-Government Findings Report, GOV.UK, published 2 June 2025 (trial 30 September to 31 December 2024; 20,000 licences; 7,115 survey responses; self-reported): [gov.uk](https://www.gov.uk/government/publications/microsoft-365-copilot-experiment-cross-government-findings-report/microsoft-365-copilot-experiment-cross-government-findings-report-html)
- Anders Humlum and Emilie Vestergaard, Large Language Models, Small Labor Market Effects, Becker Friedman Institute for Economics, University of Chicago, Working Paper 2025-56, May 2025 (Denmark; surveys in November to December 2023 and 2024; about 25,000 valid responses per round): [bfi.uchicago.edu](https://bfi.uchicago.edu/wp-content/uploads/2025/04/BFI_WP_2025-56-3.pdf)
- Workday, New Workday Research: Companies Are Leaving AI Gains on the Table, press release, 14 January 2026 (3,200 full-time AI users at organizations with $100 million or more in revenue; fielded in November 2025 by Hanover Research): [Workday newsroom](https://en-sg.newsroom.workday.com/2026-01-14-New-Workday-Research-Companies-Are-Leaving-AI-Gains-on-the-Table)

*Dan Fitzpatrick, The AI Educator, is a Forbes contributor, international keynote speaker and bestselling author who has trained more than 150,000 people across 30+ countries. [About Dan](https://theaieducator.io/about).*

## Frequently asked questions

### What should a small business do with the time AI saves?

Decide before the time is saved. Choose one destination per job: Grow the work that earns, Fix a job that keeps slipping, Build a skill that will matter more, or Return the hour to the person. Name an owner and a review date, or the hours drift into more of the same work.

### How much time does AI actually save at work?

Reported savings are modest and self-reported. A UK government trial found an average of 26 minutes a day among 7,115 survey respondents, and a Danish study found 2.8% of work hours. Workday found nearly 40% of savings lost to rework, so check your own numbers before you count on them.

### Why does the time saved by AI disappear?

It has no destination. Studies show saved time mostly goes into other job tasks, rework on the output, or extra workload, and nobody chose any of those. When no one decides where an hour goes, the business absorbs it into the work it was already doing.

### Should staff keep the time AI saves them?

Sometimes, as the Return destination, when someone is stretched or the saving came from their own effort. It should be a choice, not a default. Tell the person what will happen to the hours. If finding a saving only brings more work, people stop mentioning savings.

### How do you measure whether AI time savings are real?

Take a before number for one named job, then check the same job after four weeks. Ask how long it takes, how often it comes back for correction, and where the hours went. A saving that returns as rework is a transfer of work, not a saving.

### Is it a good idea to use AI savings to avoid hiring?

Only after you test it. Before you skip a hire, write down what the role actually did each week, which tasks AI can now do with a person checking, and what only a person in that seat can own. Saved hours rarely add up to a whole role.

---
Source: [What Should Your Business Do With the Time AI Saves?](https://business.theaieducator.io/posts/what-to-do-with-time-ai-saves)
Publisher: [The AI Educator](https://theaieducator.io)