‘It dramatically increases the time HR has to spend on these things'

Picture an employee who would like a fourth week of vacation. Instead of raising it with their manager, they send HR a long email accusing the company of breaking the law and warning of regulatory complaints. It might read: "You violated the ESA for vacation time. Your entire company is on alert."

It's a scenario that Randy Ai, principal lawyer at Randy Ai Law Office in Toronto, uses to show what can happen when an employee runs a simple request through an AI chatbot.

More employees are turning to AI to check whether they have a valid complaint and then to write it up, according to Ai and two other employment lawyers speaking to Canadian HR Reporter.

For HR, that can mean more complaints and longer, more complicated complaints that may, ultimately, have no bearing. So, how should they be handled?

More complaints, longer complaints

Dan Wilband, a labour and employment lawyer with VanBuskirk Law in Saint John, N.B., agrees the issue comes up more frequently. At a conference on administrative tribunals he attended recently, there was a lot of discussion about tribunals and courts receiving “greatly increased numbers” of submissions that were obviously produced with AI.

“That spills over quite naturally into workplaces where I think employers are experiencing an increased volume of probably AI-generated complaints,” he says, and the same is true for legal counsel, where the phenomenon is “very quickly proliferating.”

Part of the burden is length. Wilband describes "the kitchen sink complaints that show up as dozens of pages or even hundreds of pages sometimes alleging everything under the sun like harassment, discrimination, reprisal, constructive dismissal” which “creates a great burden on HR” along with lawyers and tribunals.

“This can be a great drain on resources,” he says.

If people don’t know how to use the tool properly, it can give them a “very confident-sounding set of expectations,” says Wilband.

"For HR, that means that things can get harder to resolve if the AI tool told somebody to expect an outcome that's really just not realistic in the circumstance.”

The wrong label can also get attached, he says, such as constructive dismissal.

Much like patients consulted the internet before seeing the doctor, employees are consulting AI before seeing HR or a lawyer, says Lai-King Hum, principal and senior lawyer at the Hum Law Firm in Toronto.

She says the shape of complaints is changing too. A harassment complaint used to be a presentation of facts: who harassed the employee, when it happened and that they want to make a formal complaint.

"Now, what you're going to see is ‘I was harassed by so-and-so, this is what the law says on what harassment is, and that's what happened to me, and it happened on this date,’ etc.," she says.

"By itself, there's nothing wrong with it as such, but it creates a lot of work… It dramatically increases the time that HR has to spend on these things.”

False confidence from AI prompts

Ai says employment lawyers are seeing AI-generated inquiries, with an exaggerated idea of how much their case is worth, along with self-represented individuals using genAI for workplace disputes.

“These are not legally educated people but they’re using AI to essentially act as a pseudo-lawyer.”

He says the effort needed for an employee to file has all but disappeared. Someone can describe an incident to a chatbot by voice, and get back a five-page discrimination complaint with perfect grammar, plus instructions on where to file, he says.

“A complaint has become cheap now… The barrier to entry for grievers, for complainants, for plaintiffs has dropped to probably the lowest point ever that I've seen in my 16 years of practice."

Ai says the most concerning pattern he sees is clients who have already retained a lawyer and then push back with a memo from ChatGPT saying their damages are worth more, or that a different legal strategy should be taken. They forget that litigation involves opposing counsel, uncertainty, costs and the employer's own arguments, he says.

"This is a huge struggle right now within the legal community, in that clients are slowly siphoning authority away from legal counsel to the AI models,” says Ai.

Hallucinated case law is a real problem, Ai says, but the bigger risk as the models improve is subtler: a false sense of confidence about damages, strategy and what a judge would decide. A client can say they have been hurt and be told they are entitled to $500,000, without any analysis of whether the conduct meets the threshold for bad faith, whether it is medically supported or whether it is discrimination under the Human Rights Code.

"The LLMs come across as very confident — you never have a sheepish answer from ChatGPT." Users also rarely ask for the other side, he says. "They never go to ChatGPT and say, ‘Please provide me with a counterargument as to why this could be wrong.’"

Polished isn't the same as credible

A document that looks polished isn't necessarily credible, says Wilband. In workplace investigations, employees sometimes use AI to help prepare their written statements. There are documented cases, he says, where it became clear in the interview that AI had invented symptoms or misrepresented medical findings that the employee couldn't explain.

"I would say HR people doing investigations should probably make sure that they're relying on their interviews with people,” he says.

Wilband also points to AI-generated medical notes and doctored text-message screenshots: "Arbitrators and courts have always treated those as serious misconduct, but these days AI just makes this easier to produce and harder to identify."

AI: power tool in any hands

However, none of the lawyers sees the trend as entirely negative.

Hum says AI has to some extent levelled the playing field for individuals by making basic legal information easier to access, and it can make employees feel more emboldened.

"People used to be afraid to report sexual harassment or other types of harassment," she says, and AI may help them come forward.

But that it is “fraught with risks” because the information may not be reliable, she says, pointing to a line from Alexander Pope whom she studied as an undergraduate: "A little learning is a dangerous thing. Drink deep or taste not."

"AI is a wonderful and great tool if you have some expertise in the area already, so you can detect, ‘Oh, this is not right, this is helpful,’" says Hum. "But if you have no idea, it just complicates things."

Wilband compares AI to a power tool: "It can do great things if you know how to use it and what it's for. It can also do terrible things if you don't."

In some cases, AI can help employees express their sense of unfairness with greater clarity, particularly employees working in a second language, neurodivergent employees, or people dealing with cases that involve a lot of documents, he says.

"I think an employee can now more easily show up with a properly framed accommodation request or harassment complaint that might, in some cases, allow it to be resolved earlier because it can be presented in a clear way from the outset.”

Ai agrees that the tools give a huge advantage to people who are less educated or less sophisticated, "because now they're able to generate very sophisticated documents using very sophisticated language that they previously just could not generate."

Talking human to human

All three lawyers say an AI-generated complaint still must be dealt with. "HR should not ignore it. Definitely not," says Hum. The AI may make it harder to figure out exactly what the employee is complaining about, she says, "but you do have to address it."

Wilband compares it to a complaint that arrives handwritten and poorly articulated: "I don’t think anybody should discount a complaint just because it looks like it was machine written.”

In most cases, a good approach is to meet the employee, ask them to explain the concern in their own words, confirm the issues in writing and discuss it in person, he says. The employee may have used AI to clearly express something they understand well, or they may be only partially aware of what's in the document.

“Those conversations are more important than ever to make sure that the right issues are being addressed,” says Wilband.

Ai says talking it through with the employee often reveals something far simpler than a complicated AI version, such as a person wanting four weeks instead of three weeks’ vacation.

"The biggest problem here is, from an HR perspective, is what is the employee’s actual intent? What do they actually want? Because the AI communication is going to give you less information than if they actually sent it themselves," he says.

Don't answer AI with AI

While it may be tempting for HR to use AI in response to a five-page complaint, Wilband says that’s probably not a good approach.

HR should instead "try and use it as a gateway to a conversation rather than an endless feedback loop of machines talking to machines” and hold their own use of AI to the same standard they expect from employees, he says.

Hum is more open to HR using AI to summarize lengthy complaints, as long as people check the work.

"If HR is going to do it, it is simply a first step and it should not be the final product. You need competent people, experienced people in HR to do the work,” she says, adding that paid products trained in a more controlled environment are more reliable than free tools.

Privacy considerations

If HR does turn to AI, privacy is a further concern. Wilband raises the question of either employees or HR teams pasting confidential personal information or health information into AI tools without the proper privacy checks. He says employers can introduce acceptable use policies covering confidential information and workplace disputes, and HR should know the privacy settings and internal policies that apply.

Hum advises against using free services, where privacy isn't protected, for this kind of material. But she says the key is to be prepared as we’re going to be more and more inundated with AI-generated content.

“You have to be savvy about using it as a tool, and be very, very careful that you’re circumscribing with appropriate confidentiality and privacy guards.”

For Ai, the effect is two-sided: "It's leveled the playing field for certain people, but certainly, at the same time, it's created a huge mess."