Employees are using AI without telling their bosses — and it seems to be paying off, according to the Financial Times. The salmon-backed newspaper published an opinion piece claiming that using AI “on the sly” might actually pay off, citing a recent study conducted by a team of academic researchers that concluded that it is possible to “get ahead” at work by utilizing AI — as long as your superiors remain in the dark, and can’t tell that your output is AI-driven.

Let’s play a guessing game. 130 middle managers at anonymous consulting firms were asked to assess a set of client briefs compiled by two junior analysts — some were written with the assistance of ChatGPT, and the others were not. The kicker? The managers could not effectively tell which was which.

The experiment yielded a few interesting results: In the AI-assisted briefs, 77% of the managers were able to correctly guess that ChatGPT had been used. But at the same time, 73% of them incorrectly guessed that ChatGPT was used when they were presented with human-sourced briefs. Even when they were corrected, 44% of the managers insisted that the human briefs were written with help from AI. AI-assisted briefs also scored a whopping 10% higher… until the managers realized that AI had been used. Once aware of the usage of AI, the managers decreased their quality rating for the briefs, perhaps under the impression that it had taken their junior analysts less effort to complete.

Here’s the takeaway: Unless employers explicitly support it, employees who use AI may be inclined to keep it on the down-low under the threat of potential bias. In May 2024, EnterpriseAM reported that 52% of employees who use AI at work are hesitant to disclose the fact, worrying that their use of the tool will make their superiors believe they were less productive. This concern seems to be corroborated by the results of the study reported by FT.

Things can get very ugly very fast. “The absence of a corporate-wide policy would result in agency problems arising from the individual use of ChatGPT, […] exposing the firm to multiple and diverse risks,” the study concludes, noting that employees keeping AI use a secret are putting the company at risk of security breaches. Any information being fed into AI is no longer confidential, and can be used in the generation of results for people outside the company, including competitors. The solution? Embrace AI in the workplace — but with clear terms.

While many companies remain hesitant to embrace AI, — given the hallucinations AI models tend to have every now and then — some have been quick to adapt. Shoosmiths — a British law firm — asked its 1.5k employees in April to utilize Copilot for a combined 1mn times throughout the current financial year to unlock a GBP 1mn bonus pot that would be distributed amongst the employees. “This initiative is about creating a culture where everyone plays a role in embracing change and using technology to raise the bar on client service,” said Shoosmiths CEO David Jackson in a press release.

Tags: