Employers are racing to catch up as workers experiment with AI without knowing the risks
Businesses need to continually monitor output from AI tools or risk liability, says legal expert
Author of the article: Marisa Coulton Published Jul 04, 2023
Employees are rushing to experiment with artificial intelligence tools that can help them do their jobs more efficiently, but in the process they might be putting their employers at risk, according to a new report by KPMG Canada.
Interest in artificial intelligence has exploded in recent months, said Zoe Willis, who leads KPMG Canada’s data and digital practice. Though the technology has existed for two decades, it has only recently become accessible to the public through tools such as ChatGPT, the AI chatbot released in November. Some companies have been caught off-guard by employees’ sudden adoption of AI.
“Businesses just haven’t been ready,” Willis said. “They don’t have the policies or the frameworks in place to actually understand how employees and people at work can use it in a safe and responsible way.”
Willis said that while employees are trying to do things “faster and smarter,” a lack of understanding of the technology and its limitations may lead to problems — if, for example, the responses from chatbots are not checked for context and accuracy “because it’s not 100 per cent factual what you get back.”
Some workers have also been putting sensitive information into ChatGPT, such as private company financial data, information about customers and clients and proprietary company data. Since the chatbot is open-source, that information could become available to individuals outside the company, Willis said.
“At the end of the day, there is risk associated with using AI, but it’s also an opportunity,” said Robert Piasentin, a partner at McMillan LLP and national chair of their technology law group. “If businesses can get on top of it and understand how the AI tools can be used, and how the risk can be mitigated in a reasonable way, there is a lot of potential for businesses to be much more efficient and way more profitable.”
For example, AI could be especially useful for a lawyer that is trying to negotiate a merger or acquisition, Piasentin said. The lawyer might need to review thousands of documents and flag potential risks by hand.
“If you can use an AI bot to take that task and do it a lot more quickly than an individual can, and still produce high quality results, businesses will be saving a lot of money,” Piasentin said.
But such work is high stakes. Companies in a regulated environment who go offside could be hit by lawsuits or find themselves shut down entirely, Piasentin said.
AI is a powerful tool, but it’s still just a tool, he said. It requires supervision.
“Someone needs to exercise some level of judgment to ensure (the AI is) capturing the right information,” he said. “Is the tool working the way you expect it to, so that you’re not missing something down the line?”
A business might need to hire a specific person or team who understands how the bot has learned, and what data set it has used, in order to periodically assess the accuracy of its output.
However, the cost of hiring these specialists might exceed the money a business hoped to save by using the AI in the first place.
The KPMG report advises businesses to take a proactive approach and develop an AI policy before they start using the tools, not after.
“What the report is showing is that (employees) are using it anyway,” Willis said.
While some workers have jumped feet-first into the world of AI, others are more wary. Some fret that the technology will take their jobs.
“I really don’t think it’s going to replace humans, I think it’s actually going to empower humans,” Willis said. “It can’t make you a cup of coffee. It can’t do a whole bunch of things…. It’s never going to replace humans, to a certain extent, but I think it can empower.”
Nicholas Walker, an English professor and chief executive of ConverSolo, an AI-powered software, echoed that AI can, indeed, make employees better at what they do.
“Artificial intelligence is a game-changer in education,” Walker said.
He said his company’s technology can be used for training purposes, allowing a worker, for example, to interact with an AI that has adopted a specific persona — such as an angry or distracted customer — to practice likely scenarios.
“We can create these kinds of role plays that are meaningful, that are supported with detailed feedback and provide the repetitions that learners need to remember the lesson,” Walker said.
Whether businesses decide to implement a tool like ConverSolo, or stick to more traditional training programs, AI will be present in the workplace, no matter what.
“I do think the rabbit is out of the hat,” Willis said. “(It) is here.”
• Email: mcoulton@postmedia.com | Twitter: marisacoulton