AI was meant to cut costs, but now it’s creating expensive problems

Artificial intelligence, the tech world’s most popular new concept, was embraced by companies hoping to reduce staff and cut costs. However, many firms that rushed to adopt this technology are now re-hiring people to correct the errors made by AI, spending a fortune in the process.

According to a report by the BBC, a new industry has emerged for software engineers and writers hired specifically to fix the mistakes made by artificial intelligence.

The experience of Sarah Skidd, a product marketing manager in Arizona, is a striking example of this trend. In May, a content agency contacted Skidd after the website copy generated by AI for a client in the hospitality sector failed to meet expectations. Skidd described the text as “very obviously written by AI, basic and uninteresting.” The company noted, “It was very bland when it needed to drive sales and spark curiosity.”

Skidd spent 20 hours rewriting the text from scratch, billing the agency $2,000 for her work at a rate of $100 per hour.

Skidd is not worried about AI taking her job; on the contrary, she says this trend has brought her more work. Speaking to the BBC, Skidd said, “Maybe I’m being naive, but if you’re very good at your job, you won’t have a problem.”

Like Skidd, many writers are now being hired not to create content from scratch, but to correct the errors in AI-generated text. This trend comes as AI tools like ChatGPT and Google Gemini gain popularity with the promise of optimizing workflows and cutting costs. According to a recent survey by the UK’s Federation of Small Businesses, 35% of small firms plan to increase their use of AI within the next two years.

However, these experiences demonstrate that AI still has a long way to go before it can match human standards.

Sophie Warner, co-founder of the Hampshire-based digital marketing agency Create Designs, says that over the last six to eight months, she has seen a surge in clients needing to fix problems created by AI. Warner explained, “Previously, clients would contact us when they had problems with their sites or wanted to add new functionality. Now, they go to ChatGPT first.”

However, adding code generated by ChatGPT can make websites prone to crashing and vulnerable to cyberattacks. In one case Warner described, a client asked ChatGPT how to update an event page. Warner noted that while the task would have taken only 15 minutes to complete manually, the AI-generated code crashed the website. This resulted in a three-day outage and a recovery cost of approximately £360 for the business.

Warner added, “We often have to charge a review fee to diagnose what went wrong because clients are reluctant to admit their mistake. The process of fixing these errors takes much longer than it would have if they had consulted professionals from the beginning.”