Employees use AI-saved time to correct AI results

According to current studies, the use of generative AI in companies is a double-edged sword: on the one hand, employees report time savings with AI use; on the other hand, significantly more time is spent correcting flawed AI results.

A survey by HR and financial software specialist Workday found that while around 85 percent of surveyed employees could save one to seven hours per week with AI. A large portion of the time saved – up to 40 percent – is spent on rework of faulty AI content. With all the error corrections, rewriting content, and reviewing results, Workday reports that only 14 percent of respondents achieve a consistently clear, positive net result with AI.

Often, there is also a lack of training in this field. While two-thirds of managers would prioritize this, only 37 percent of employees who have to do most of the rework stated they also have access to training. Furthermore, savings through AI only benefit employees to a limited extent: around 32 percent of companies simply increased the workload – and left it up to employees to deal with AI themselves. According to Workday, 3200 employees worldwide who actively use AI and work for companies with at least 100 million US dollars in annual revenue were surveyed.

Several Hours of Rework Per Week

A survey by automation service Zapier among 1100 AI users in companies estimates the rework effort for poor AI content at 4.5 hours per week, which is just over half a working day. 58 percent indicated they spend three or more hours per week on it, 35 percent at least five hours, and eleven percent even ten or more hours. Only three percent indicated that AI rarely needs correction.

Nevertheless, 92 percent reported that AI makes them more productive. Around half even reported significant increases in their productivity. Only one percent said that AI actually makes them less productive. Almost three-quarters have already encountered initial negative consequences from AI blunders (74 percent). These include customer complaints, data protection and security incidents, or work results rejected by colleagues due to quality defects.

The AI correction effort varies by department. For engineers, IT professionals, and data specialists, it averages five hours of rework, and 78 percent have already encountered problems due to AI content. In finance and accounting teams, it averages 4.6 hours of AI corrections, and 85 percent have reported negative consequences due to plausible-sounding AI nonsense. Sales and customer support seem less affected, with three hours of rework and 62 percent reporting encountered difficulties.

No? Yes! Oh!

Overall, the marketing narrative seems to be shifting under the convergence of generative AI and work reality. Instead of promises of unlimited productivity and employee replacement, the focus is increasingly on "hard work" and change management. Companies would have to undertake this if they wanted to reap the benefits of the technology. This was recently heard from Microsoft CEO Satya Nadella during his appearance at the World Economic Forum in Davos, among other things.

Measurable value from AI implementation has not yet been achieved by the majority of companies, as various studies by consulting firms Deloitte, PwC, and BCG suggest. Only around 12 percent of surveyed companies have so far achieved cost savings and value creation, PwC noted, for example. (axk)