Some employers are clear about AI, offering bonuses to staff who use it regularly and warning of possible disciplinary action if it is misused.
Others have failed to set even minimum guidelines, leaving many workers confused as to how to deploy new tools in their jobs.
The opacity surrounding company policies is resulting in staff running personal generative AI accounts in secret or discovering they have inadvertently breached the rules, according to lawyers and employment experts.
The problem stems from the early days of companies “aggressively telling people not to use AI, over fears of private data leaking”, says Joshua Wöhle, chief executive of Mindstone, an online platform that offers AI training. “It’s had a big effect on people being scared to use it.”
Even now that employers are sanctioning AI in the workplace, “the hangover [is] long-lasting”, he adds. The result is that there are “people who don’t use it at all. Or those that do are using a private account.” He warns the problem is “dramatically bigger” than many executives assume.
Jason Ross, employment partner at US law firm Wood Smith Henning & Berman, says he has been contacted by managers who are “deeply concerned” about “purposeful or merely unintentional negligence” by staff who are oblivious to risks and company guidelines. Employers’ most common complaint, he says, is of individuals using AI tools without “appropriate oversight and review”.
In the legal industry, there have been high-profile cases that have broken court rules regarding the need to disclose AI use, and have resulted in “attorney and firm embarrassment, sanctions, negative publicity, negative client outcome, exposure for professional liability, and significant reputational harm”, he says. “It can also have grave disciplinary consequences to the implicated attorney, including job loss.” A Californian court issued a fine after a lawyer generated “fake legal authority by AI sources”.
Sinead Casey, head of UK employment at law firm Linklaters, expects disciplinary issues to emerge in employment tribunals. “Clients [are] incorporating [misuse of AI] into their disciplinary policies. It’s quite early days. I’m sure it’s coming.”
To avoid cases escalating to that level, business leaders are advised to set clear policies. These typically cover “confidentiality, risks of bias and discrimination, the need for human engagement and review, and data privacy and security issues”, says Casey. They should also make clear which AI tools are permitted, she adds, whether the employee should be disclosing to managers when tools are used to perform tasks, and make clear a breach may lead to disciplinary action up to and including dismissal.
Ross adds that broadly, typical corporate policies mandate “baseline human involvement, not total AI function displacement”. They should “dictate controls on what can and cannot be input into AI systems, and what AI tools or systems employees are and are not allowed to use”.
Others suggest reinforcing policies with a positive tone from senior managers. “If the executive team does it and highlights the use of it in a way that is positive, it trickles through,” says Wöhle. Companies should also designate trained staff to provide examples of use cases and answer questions on policies, as the technology and applications evolve.
Culture is critical in encouraging disclosure of gen AI assistance, adds Wöhle. “Is AI something celebrated or something to hide? If you use AI, do people feel proud of using AI to get a result? Or do people hide it as they fear they will look like they have been cheating? If it’s unclear to people whether they will be celebrated or called out for it then one will lead to better adoption and [disclosure and] one won’t.”
This year, a global survey by KPMG and the University of Melbourne of 48,340 professionals found 44 per cent of employees contravene organisations’ policies and guidelines on AI, and 61 per cent hide their use of AI tools at work with more than half passing off AI content as their own.
Niale Cleobury, global AI workforce lead at KPMG, says: “We’ve got to be positive [and let] people feel they can be experimental.” This might mean discussing in a team meeting how AI was used and sharing ideas, for example. “I don’t want to police it, but I want to know how they’ve used it.”
The challenge for employers is that they are trying to write static rules for dynamic technology, says Rohan Sathe, chief executive of Nightfall AI, a cloud data protection platform. He believes employees are “not trying to hide” AI use but “simply don’t know it’s a problem” because most organisations have not communicated clear policies.
He advises employers to provide real-time contextual guidance over messaging platforms such as Slack, email or video calls. “A policy written in March is outdated by November.”
Casey agrees the pace of development means AI policies cannot be done once to then “stick in a drawer”. “This is unprecedented. There [isn’t] anything comparable in terms of the meteoric rise of gen AI but also [its implications] on legal and HR.”
In spite of the potential risks, 19 per cent of respondents to Deloitte’s Digital Consumer Trends research said their company did not have a policy or guidance about gen AI and 14 per cent did not know whether their company had a policy. A survey by WalkMe, an SAP company, found that employees were using AI tools in secret because the company-sanctioned options were poor quality.
Some companies set bonuses to encourage open and appropriate AI use.
One organisation created a leaderboard, offering $10,000 to the employee who made the highest number of quality prompts, as well as the most innovative use, which peers voted on. “The engagement was off the charts,” Wöhle says.
Emma Grede, co-founder of Skims, the shapewear company, and the denim brand, Good American, has discussed incentivising staff with a “bonus system” to use AI to improve productivity.
In April, law firm Shoosmiths created a £1mn bonus to share among employees if they reached a target of 1mn prompts over the year. They are set to meet it several months early.
Staff can use Microsoft’s Copilot to summarise and polish documents, as well as performance reviews but are banned from turning to AI for any specific legal queries.
Tony Randle, a partner at the firm, says: “It’s not trained to be a lawyer. It’s very credible but it absolutely doesn’t have it. We encourage people to use it for clarity, sentiment and brevity.”
The company is also exploring legal platforms designed to help lawyers with their tasks. But it does have some clear restrictions in place. Aware that responsible use of AI extends to being mindful of the environment, Shoosmiths has banned staff from using the technology to produce images. “The generation of images by AI produces 10 times the energy a word-based output would require. We would not want people to burn through energy.”
Read the full article here
