# Will Clients Use AI Instead of Hiring Your Business?

Canonical URL: https://business.theaieducator.io/posts/clients-using-ai-instead-of-hiring-you
Publication: Dan Fitzpatrick Insights
Author: Dan Fitzpatrick
Topic: AI and the Future of Work
Published: 2026-11-12T16:40:18.000Z
Modified: 2026-11-12T16:40:19.827Z

Clients are taking back first-draft work they can now do with AI and still paying for work that is finished, judged and answered for. Here is what three studies measured, and a three-question test to run on everything you sell.

## In brief

Some clients will use AI instead of hiring you, but only for work they can do passably, judge for themselves and live with if it is wrong. Run the three-question Good-Enough Test on each service, stop defending the draft, and move exposed work to a finished job or a paid check.

## Key takeaways

- A study of 1,388,711 job posts on a global freelancing platform found posts for writing, software and engineering work fell 21% more than manual-intensive work within eight months of ChatGPT's release, with writing down 30.37%.
- The jobs that remained in those categories asked for more skills, carried budgets 5.71% higher and drew 8.57% more bids each.
- A November 2026 OpenAI and Upwork report found content writing made up 14.5% of sampled ChatGPT conversations but about 1.3% of Upwork job posts, and that job posts involving AI carried fixed-price budgets about 52% higher.
- Clio's 2026 Legal Trends Report found 46% of US consumers would consult AI before contacting a lawyer, yet only 26% are comfortable with AI making decisions.
- The Good-Enough Test asks three questions of each service: could a client get a passable version from AI, could they tell good from bad without you, and could they live with it if it were wrong?
- Three yeses mark work a client can stop buying; any no marks what they still pay for, which is framing, judgment and ownership.
- Stop defending the draft: fold exposed work into a finished job you stand behind, or sell the check on what the client drafts, and tell clients how you use AI yourself.

Some will, for some of the work. The evidence so far says clients use AI in place of a supplier for jobs they can do passably and check for themselves, mostly first-draft writing, and still pay for work that has to be finished, judged and answered for. Your task this week is to find out which of your services sit on which side.

You have probably seen the first sign already. A client who used to send a brief now sends a draft and asks you to "tidy it up." Another has gone quiet on the monthly retainer. Nobody has said the word AI, and nobody needs to.

The worry is reasonable, and it is also too broad to act on. "AI will replace service businesses" tells you nothing about Monday. What clients are taking back is narrower than that, and it has been measured.

*Sources checked 12 November 2026.*

## What does the evidence say about clients using AI instead of a supplier?

The first draft is leaving and the finished job is not: that is the pattern in all three sources I read this week.

**The marketplace study.** Ozge Demirci, Jonas Hannane and Xinrong Zhu studied 1,388,711 job posts on a leading global freelancing platform, across 61 countries, from July 2021 to July 2023. Within eight months of ChatGPT's release, posts for writing, software and engineering work fell 21% more than posts for manual-intensive work such as data entry, audio and video. Writing fell most, by 30.37%. Image creation posts fell 17% after the image generators arrived.

The detail owners miss is what happened to the jobs that stayed. Among employers who posted both before and after, the remaining writing and coding jobs asked for more skills (2.18% more skill tags), carried budgets 5.71% higher and drew 8.57% more bids each. The work that survived was harder, better paid and more crowded. The paper appeared in Management Science in 2025; the figures here are from the August 2024 working paper, which is the version I read. ([Demirci, Hannane and Zhu, CESifo Working Paper 11276](https://www.ifo.de/DocDL/cesifo1_wp11276.pdf))

**The report from this month.** In November 2026, researchers at OpenAI and Upwork compared what people do in ChatGPT with what clients hire freelancers to do on Upwork, a global marketplace. They classified 36,896 sampled ChatGPT conversations from January 2025 to June 2026. Content writing made up 14.5% of those conversations and about 1.3% of Upwork job posts in the first half of 2026. From early 2023 to late 2025, content-writing posts fell at more than twice the rate of posts overall, while video and animation grew by about a third relative to the marketplace. Across 63 categories of work, 43 fit one pattern: what people do heavily in ChatGPT lost share of freelance hiring, and what they do little of there gained it.

The same report found that job posts involving AI carried fixed-price budgets about 52% higher than other posts in the same category and month, and 41% more hours in the first 90 days of hourly contracts, at almost the same hourly rate. Clients were buying bigger jobs, not pricier hours. Two cautions: the authors work for the two companies, and the report says its datasets are not matched, do not represent the labor market as a whole, and do not show cause. ([Chin and others, The Rebundling of Freelance Work in the AI Age, OpenAI and Upwork, November 2026](https://cdn.openai.com/pdf/openai-upwork-full-report.pdf))

**The consumer side.** Clio's Legal Trends Report, released on 26 October 2026, found that nearly one in four US adults has used AI to answer a legal question and 46% say they would consult AI before contacting a lawyer. Yet only 26% are comfortable with AI making decisions, and 6% with its use in high-stakes personal matters. The release gives no sample size, and Clio sells software to law firms, so read it as a signal. ([Clio press release, 26 October 2026](https://finance.yahoo.com/technology/ai/articles/clios-2026-legal-trends-report-133000373.html))

If your clients are small businesses, they are doing the same. Simply Business surveyed 1,842 UK small business owners between 30 July and 7 August 2026: 47% now use AI, up from 22% a year earlier, and 63% of those users create content with it. ([Simply Business, 8 September 2026](https://www.simplybusiness.co.uk/knowledge/business-news/small-business-ai-usage-doubles/))

None of these sources measures how much revenue a small firm loses when clients serve themselves. I could not find one that does. What follows is my reading, not their finding: clients take back the part of a job they can do passably and check for themselves, and what they still buy gets bigger and harder.

## Which of your services are most exposed?

The exposed services are the ones where a client can make a passable version, tell whether it is any good, and live with it if it is wrong.

That gives you a test for each thing you sell.

The Good-Enough Test is three questions to ask of each service you sell: Could a client now get a passable version from an AI assistant in an afternoon? Could they tell a good version from a bad one without you? And if it turned out wrong, could they live with it? Three yeses mean the client can stop buying it whenever they choose. Any no marks what they are still paying you for.

I suggest it as a way of thinking. It is not established research, and the three questions are mine.

Take a hypothetical 12-person marketing agency and run its five lines of work through the test.

- **Monthly blog posts.** Yes, yes, yes. The client can draft them, can judge them, and a dull post costs little. This line is at risk.
- **Social captions.** Three yeses again.
- **Campaign strategy.** A client can get a plausible strategy from an assistant in an hour. Most cannot tell a sound one from a confident one. The second answer is no, and that no is the service.
- **Managing the ad account.** The client could try. A mistake spends real money before anyone notices. The third answer is no.
- **A product video shoot.** The first answer is no. Somebody has to turn up with a camera.

Two of the five are exposed, and they are probably the two the agency finds easiest to deliver. That is not a coincidence. Work that is easy for you to hand to AI is easy for your client to hand to AI.

## What should you do about the exposed work?

You have three ways to respond to a three-yes service, and the evidence favors two of them.

| Approach | What the sources say | Fits when | Breaks when |
|---|---|---|---|
| Defend the draft: keep selling the same thing, faster or for less | Posts for this kind of work fell 21% more than manual work, and each remaining job drew 8.57% more bids (Demirci and others) | Your clients have not tried AI yet | They try it |
| Sell the finished job: fold the draft into a larger result you deliver and stand behind | Jobs involving AI combined more tasks and carried budgets about 52% higher, with 41% more hours (OpenAI and Upwork); the surviving jobs were more complex (Demirci and others) | You can deliver and check the whole result | You take on parts you cannot judge |
| Sell the check: the client drafts, you review, correct and put your name to it | Only 26% of US consumers are comfortable with AI making decisions, 6% in high-stakes matters; 82% want to know when their lawyer uses AI (Clio) | An error is costly or the work is regulated | The client cannot see what your review adds |

My verdict: stop defending the draft. Move each exposed service to the finished job or to the check, and say so to the client before they raise it. The third row rests on consumer attitudes in one profession; no source I read measures a review-only service, so treat it as my inference.

For the hypothetical agency, blog posts stop being a line on the invoice. They become one part of "we run your content and report what it brought in," or they become "you draft, we edit, fact-check and publish under a standard we both sign."

## Why does the Scarcity Rule explain what clients still pay for?

Clients keep paying for the skills that get scarcer as AI gets better, and those are the three the test is looking for.

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 it: framing, judgment and ownership.

Each maps onto the test. A client with an assistant can produce the thing, which is why the first question so often comes back yes. What the assistant does not give them is framing, knowing which thing is worth producing. The second question is judgment: can they tell good from bad? The third is ownership: who answers when it is wrong? A service built on those three does not get cheaper to copy each time the tools improve. A service built on producing the draft does.

If AI is saving your own team time, this is where to spend it. [The Dividend Decision](https://business.theaieducator.io/posts/what-to-do-with-time-ai-saves) calls that destination Build.

## Is this happening to every kind of business?

No. The measured shift is in digital work bought at arm's length, and it weakens as work becomes physical, regulated or dependent on a relationship.

Two of the sources are online freelance marketplaces, where a client can stop hiring with one click and nobody phones to ask why. A retainer with a firm you have known for six years has more friction, which buys you time, not immunity. If your business works with its hands, the first question is no for most of what you sell, and your exposure is in the paperwork around the job: see [where AI fits in trades and construction](https://business.theaieducator.io/posts/ai-for-trades-and-construction).

There is a second route by which clients use AI, and it applies to everyone. Before they contact a supplier, many now ask an assistant who to hire. [Getting your business recommended by AI](https://business.theaieducator.io/posts/get-business-recommended-by-ai) covers that.

## The Mistake I See Most Often

The mistake is hiding your own use of AI while still selling the draft as if it were handmade.

Across the business teams I train, the owners most worried about clients leaving are often the ones saying least to those clients. They use AI to produce the first version in a fraction of the time, send it out under the old description, and hope nobody asks. Then a client makes something similar on a Sunday evening and begins to wonder what the invoice was for.

I was a teacher and an assistant headteacher before I worked with businesses, and schools met this problem early. Once a student can produce a passable essay in a minute, the essay stops being evidence of anything, and the value moves to the conversation about it. Your draft is in the same position.

So tell them. Say where you use AI, what you check, and what you stand behind. [Whether to tell customers you use AI](https://business.theaieducator.io/posts/ai-disclosure-small-business) covers how, and [the Name-On-It Rule](https://business.theaieducator.io/posts/questions-to-ask-suppliers-who-use-ai) lists the five answers a careful client will want from any supplier, including you. If you are unsure who owns what you produce with AI, [the Fingerprint Test](https://business.theaieducator.io/posts/ai-generated-marketing-content-ownership) helps.

## What can you do this week?

You can run the Good-Enough Test on your own invoices in an afternoon.

1. List what you billed for in the last three months, by type of work, not by client.
2. For each type, answer the three questions with the person who delivers it. Write yes or no, and be harder on yourself than feels fair.
3. Ask three clients one question: "What have you started doing yourselves with AI this year?" Most will tell you if you ask plainly.
4. For every three-yes service, choose the finished job or the check, and write the one sentence you would say to a client about it.
5. Tell your clients how you use AI before they find out another way.
6. Put a date in the calendar 90 days out and run the test again. The tools improve, and a no on the first question does not stay a no.

If the test shows that most of what you sell is three yeses, that is a strategy question, not a marketing one. It is better found now than in a renewal meeting. And if you conclude it is [too late to start with AI yourself](https://business.theaieducator.io/posts/too-late-to-start-using-ai-small-business), the evidence says otherwise.

## Where to go next

If you are planning a members' event, a client day or a leadership offsite on what AI does to the work people pay for, that is what [my business keynotes](https://theaieducator.io/business-keynotes?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=clients-using-ai-instead-of-hiring-you) are for. To see where your own business stands first, the free [Business AI Readiness Scorecard](https://theaieducator.io/business-ai-readiness?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=clients-using-ai-instead-of-hiring-you) takes about five minutes, and if your team needs to get better at the checking and judging clients still pay for, [AI training for teams](https://theaieducator.io/ai-training-for-business?utm_source=business.theaieducator.io&utm_medium=referral&utm_campaign=clients-using-ai-instead-of-hiring-you) is where I would start.

## Sources and further reading

- Ozge Demirci, Jonas Hannane and Xinrong Zhu, Who Is AI Replacing? The Impact of Generative AI on Online Freelancing Platforms, CESifo Working Paper 11276, August 2024; published in Management Science, 2025 (1,388,711 job posts on an undisclosed global platform, 61 countries, July 2021 to July 2023): [ifo.de](https://www.ifo.de/DocDL/cesifo1_wp11276.pdf)
- Caroline Chin, Shagun Kala, Nadav Kunievsky, Teng Liu, Alex Martin Richmond and Carl Shan, The Rebundling of Freelance Work in the AI Age, OpenAI and Upwork, November 2026 (36,896 sampled ChatGPT conversations, January 2025 to June 2026; Upwork Marketplace job posts, global; descriptive, not causal): [OpenAI](https://cdn.openai.com/pdf/openai-upwork-full-report.pdf)
- Clio, Clio's 2026 Legal Trends Report reveals record gains in billable time as AI reshapes law firm economics, press release, 26 October 2026 (surveys of US consumers and legal professionals; sample sizes not given in the release): [press release](https://finance.yahoo.com/technology/ai/articles/clios-2026-legal-trends-report-133000373.html)
- Simply Business, AI use among UK small businesses more than doubles in a year, 8 September 2026 (1,842 UK small business owners, surveyed 30 July to 7 August 2026): [Simply Business](https://www.simplybusiness.co.uk/knowledge/business-news/small-business-ai-usage-doubles/)

*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

### Will clients use AI instead of hiring my business?

Some will, for some of the work. The evidence so far shows clients using AI in place of a supplier for first-draft work they can do passably and check themselves, mostly writing. They keep paying for work that has to be finished, judged and answered for when it goes wrong.

### Which services are most at risk from clients using AI?

The services most at risk are those where a client can make a passable version, tell whether it is good, and live with it if it is wrong. In the studies, writing fell most, followed by software and image creation. Work needing physical presence, specialist judgment or accountability held up better.

### What is the Good-Enough Test?

The Good-Enough Test is three questions to ask of each service you sell: could a client get a passable version from AI in an afternoon, could they tell good from bad without you, and could they live with it if it were wrong? Three yeses mean they can stop buying it.

### What should I do if clients can do my work with AI?

Stop defending the draft. Either fold that work into a larger finished result you deliver and stand behind, or offer to review, correct and sign off what the client drafts. Tell clients which you are doing before they ask, and run the test again in 90 days.

### Should I tell clients that I use AI in their work?

Yes, in most cases. In Clio's 2026 survey of US consumers, 82% wanted to know when their lawyer uses AI and 70% said they trust a lawyer more when it is disclosed. Say where you use it, what you check and what you stand behind.

### Is AI reducing demand for freelancers and agencies?

It is reducing demand for some work and raising it for other work. One study found job posts for writing and coding fell 21% more than manual work on a freelancing platform. A November 2026 report found jobs involving AI came with larger budgets and more hours.

---
Source: [Will Clients Use AI Instead of Hiring Your Business?](https://business.theaieducator.io/posts/clients-using-ai-instead-of-hiring-you)
Publisher: [The AI Educator](https://theaieducator.io)
