Why are benefits lagging in AI adoption?

Six in 10 Canadian HR professionals now use artificial intelligence (AI) tools daily or weekly, but when it comes to benefits administration, adoption drops to just 3.9 per cent – the lowest of any human resources (HR) function, according to Rise People's 2026 report, Navigating Transformative Times: The State of Canadian HR. That gap isn't a failure of ambition. It's a reflection of what's actually at stake when AI touches an employee's health data, drug coverage, or disability claim, says Kirstin Grant, Director, Total Rewards and Actuarial Consulting at KPMG in Toronto. 

"AI adoption tends to start in areas where the consequences of error are relatively low," says Grant. "In HR, third party recruiting tools, which are almost universally used by organizations, have been utilizing AI for years, so the high rate of AI use reported by the Canadian HR professionals in the survey makes perfect sense to me and is supported by what we're seeing in organizations.” 

Grant adds that benefits administration sits at the very opposite end of the risk spectrum from recruiting, contributing to the low percentage of HR professionals using AI for it. 

She points to a specific combination of factors driving that caution: "The gap is understood by looking at what's driving this risk differential – namely, the highly sensitive nature of data involved with benefit processes,” she says. “Canada's exceptionally complex laws and regulations applicable to employee benefit plan administration, the negative and potentially severe impact of errors on employees, and the number of systems owners involved all contribute to high risk of using AI in benefits administration." 

However, Grant believes that caution is a positive signal rather than a shortfall, given the data involved in benefits. "The caution and pace at which AI use in benefits administration is actually quite encouraging," she says. "I think it shows the sector's savviness and awareness of the significant risks at play, so it's not necessarily a bad thing that the survey results show this is the lowest area of AI employment – but there will be great movement in the months and years to come." 

Where the real AI adoption obstacle lies 

The barrier isn't so much AI accuracy as it is basic infrastructure, according to Daniel Drolet, Senior Partner, Group Benefits at total rewards consulting firm Normandin Beaudry in Toronto. Different systems and parties still struggle to move data between each other automatically, he says. 

"It's more the fragmented ecosystem," says Drolet. "Payroll, human resources information systems (HRIS), insurance companies – we struggle to have data transfer from one provider to the other, and automated data transfer, not just upload of files, takes a lot of time and effort.” 

He adds that it can be costly with the HRIS and payroll providers to implement data transfers, but once it's implemented, it can work well – but we're not even talking about AI, it's just basic flow of information." 

The incentive to fix this sits with carriers and providers, not individual employers, and misaligned economics may explain why it hasn't happened faster, says Drolet. "I think it's the industry who should be investing in AI to lessen the administration impact – It should not be done at the plan sponsor employer level,” he says. “And then the providers can scale, invest, and use their portfolio of clients to amortize the cost compared to an employer where they have to develop their own solution." 

On data privacy, Drolet believes that Canada's insurance regulatory framework as a source of reassurance rather than concern. "I've never seen any circumstances where someone crossed the line, used the data for wrong reasons, and tried to limit access to coverage with sensitive data," he says. "So I think it's working, we can trust the system." 

His advice for HR leaders ready to move: start with what's already licensed. "My first question would be what are you using on a day-to-day basis? What kind of license, who has access to it, and how is the data used?" he says. "Understand their ecosystem and what solutions are already in place, and then try to leverage that as much as possible." 

A tale of two benefits functions 

Parvathi Subramanyam, Senior Director of Total Rewards at 1Password in Toronto, sees the adoption gap as a split between what her team controls and what it doesn't. 

"I view the benefits world as two separate pieces," says Subramanyam. "You've got one piece of benefits administration where it's directly linked to third parties – essentially there's a lot of limitations on the AI front that isn’t from our perspective, it's the limitations from our third parties."  

She cites one striking example from her own team: "One of our third-party [providers] has to fax enrollments to carriers, as they think that that's the most secure way to send information,” she says. “They've not caught up with our times." 

On the internal side, the picture for Subramanyam’s organization looks entirely different, she says. "There's a very, very big piece of benefits work that’s internal – we're talking about standard operating procedures, process improvements, documenting operational efficiencies, data analyzing, trends, communications for internal employees," she says. "We are, I would say, at the forefront of HR, and my benefits team is using AI about 80 per cent of the time."  

According to Subramanyam, her team relies on anonymized data rather than personal information: "We do not put in personal information into AI as it relates to anything benefits-related or compensation-related." 

Closing the AI gap in benefits 

Both Grant and Subramanyam see genuine convergence ahead, though neither expects benefits to fully match other HR functions. "I do think adoption will increase significantly over time, but it will take longer and it may look different from recruiting or reporting," says Grant, adding that she doesn't expect benefits administration to become fully automated: "Benefits administration has structural characteristics that will always require greater level of oversight than these other HR processes… it will be intelligently augmented benefits administration where AI handles routine work and HR professionals focus on the controls, judgment, empathy, and employee support." 

Subramanyam agrees the ceiling depends largely on vendors catching up. "It really depends on how these third-party vendors essentially catch up from a technology perspective and how our HRIS systems catch up,” she says. “If that's taken care of, then I feel like AI adoption in the rest of benefits [within organizations] will really skyrocket because it’s within their control."