“I'm paid to fix the issues caused by AI.”

AI makes me extra money, says Sarah Skid, a product marketing manager who writes for tech companies and startups.

In May, Ms. Skid was approached by a content agency to urgently rework copies of the website created via AI generated for hospitality clients.

Instead, what was supposed to save money caused a lot of problems.

“It was the kind of copy you usually see in AI copies. It's very basic. It wasn't interesting,” Skid says.

“It was supposed to be selling it and conspiracy, but instead it was very vanilla.”

Ms Skid spent about 20 hours rewriting the copies and charged $100 (£74) per hour. Instead of making small changes, she “had to do it all over again.”

Ms Skid, who lives in Arizona, doesn't worry about companies switching to AI like ChatGpt, rather than using copywriters like her.

“Maybe I'm naive, but if you're very good, I don't think you're a problem.”

For now, she has heard the writers whose current main role is to modify copies that have been churned out by AI.

“Someone connected with me and said that's 90% of my job right now. So it's not just that I make money from those mistakes, but there are other writers.”

Ms Skid certainly doesn't mean anti-AI, she believes it could be a great resource.

“My husband and son are dyslexia and it's very difficult to write for them – anything to help someone write. It can change your life.”

Over the past few years, generative AI has taken off, and businesses have turned to systems like Openai and Google Gemini's development, transforming business practices and cutting down time and money.

A third (35%) of SMEs plan to expand their use of AI within two years, with a survey by the SMEs Federation that has increased to 60% of those aiming for rapid sales growth.

However, some companies are in a hurry and, as Skid shows, can often create more work and costs than originally intended.

Certainly, that's the experience of Sophie Warner, co-owner of Create Designs, a digital marketing agency in Hampshire, UK.

In the last six to eight months, she has seen a surge in requests for help from clients who have turned to AI for quick fixes but have come across issues.

“If a client has issues with the site or wants to introduce new features, before they send a message to us,” Warner says. “Now they're going to chat first.”

Warner said this led to clients adding code to the website proposed by ChatGpt. This has caused the website to crash and make clients vulnerable to hackers, she says.

She points to one client who says it took 15 minutes instead of manually updating the event page.

The error ultimately “we cost them about £360 and their business fell for three days.”

Warner says it will happen to bigger clients as well.

“We spend more time educating our clients about the outcomes. [of using AI].

“In many cases, you'll need to charge an investigation fee to find out what's wrong. They don't want to admit it, so the process of correcting these mistakes takes much longer than if the expert was consulted in the first place.”

Professor Feng Li, Associate Dean of Research and Innovation at Bayes Business School, says some companies are too optimistic about what AI tools can do today.

He points out that AI is known to be hallucinated – generating content that is irrelevant, constructed or inconsistent.

“Human surveillance is essential,” he says.

“We've seen companies generate low-quality website content and implement broken code that destroys critical systems.

“Unsatisfaction leads to reputational damage, unexpected costs, and even serious debt, which often requires professional rework.”

In Gujarat in northwestern India, copywriter Kashish Balot says he edited content written by AI for his US-based client to make him look more human and removed patterns of sentences that sounded like AI.

Despite the poor quality of the content, she says her clients are getting used to the speed of AI, which creates unrealistic expectations.

“AI really makes everyone think it's just a few minutes of work,” says the client, who uses Open AI's ChatGPT.

“How good a copy editing, like writing, requires consideration like AI, and takes time because it doesn't have to be curated. This is because it curates the data, so I don't really understand the nuance.”

Hype around AI has encouraged many businesses to experiment without realistic understanding of what clear goals, the right infrastructure, or what technology can offer, Professor Li says.

“For example, businesses need to assess whether they have the right data infrastructure, governance processes, and internal capabilities to support AI use. Relying on off-the-shelf tools without understanding the limitations can reduce results,” he says.

Openai says ChatGpt is useful for a wide range of tasks, but the results will depend on the model used, the experience in AI, and how prompts are written.”

It also points out that there are several versions of chatgpt.

“Each model has different capabilities to complete different tasks.”

Is Warner worried about the impact of AI?

“Yes, no,” she says. “It seems like a quick and inexpensive option, but AI rarely takes into account designs that focus on unique brand identities, target demographics, or transformations. As a result, much of the output appears to be common and can actually undermine the brand's reputation and effectiveness.”

She adds: “AI can be a useful tool, but it cannot replace the value of human expertise and context in our industry.”