Most employer-sponsored benefits packages share a structural weakness that has nothing to do with what's in them. Employees don't engage with them — and the coverage they receive rarely reflects who they actually are or where they live. That gap, according to Nick McMenemy, Mercer Marsh Benefits UK leader, is the problem AI is best positioned to solve, and the one the industry has been slowest to address.
"The biggest challenge we have around engagement is the lack of personalization and the lack of pathways through different benefits," McMenemy said in the latest episode of Marsh's Transforming Benefits with Technology vodcast series, published earlier this year.
The observation sits alongside a broader market reality. A WTW survey of 312 employers published in May 2026 found that just 20% are currently operationalizing AI within their benefits programs, even as 72% plan to embed it within two years. The gap between those two numbers - intent versus current practice - is where most of the advisory work is actually happening right now.
McMenemy and Andrew Owens, chief technology officer at Marsh Health and Benefits, both point to personalization as the clearest application for AI in the benefits space, and the one with the most direct health implications. If a plan can be designed around an individual's life stage, location, and health profile rather than a workforce average, the argument runs, employees use it more - and healthier, more engaged employees represent a quantifiable return for employers.
Owens illustrated the logic with a point about context: "If I live in the middle of the countryside, I don't want to be offered a gym. There are no gyms around. I want something alternative." The example is deliberately simple, but the underlying principle is what's driving significant investment.
The predictive dimension Owens describes goes further than enrollment navigation. He pointed to the automotive and logistics sector, where telematics data originally collected for fleet compliance is now being used to identify occupational health patterns. Flagging, for instance, whether a driver who starts late may be experiencing sleep problems, or whether frequent stops suggest physical pain. That data then informs individual occupational health conversations and, significantly, can be fed back to insurers for more accurate risk pricing, adjusting premiums to better reflect the actual health profile of a workforce.
Both speakers are consistent on what stands between the technology and its adoption: employee willingness to share personal data, and employer responsibility for how that data is used. Owens notes that organizations need to demonstrate "that their data is being used to personalize their experience and not being used for meta-analysis that they don't give permission to."
That concern is not abstract. Prudential's 2026 Benefits & Beyond study, which surveyed 3,096 US workers, found that 83% of employers want to use AI to help employees better understand their benefits, while only 58% of employees say they would use it for that purpose, and just 24% do today. The 27-point gap between employer enthusiasm and employee readiness is a direct expression of the trust problem the Marsh speakers describe.
Building that trust, Owens argues, requires an organizational posture rather than a technical fix. He calls it an "AI-first culture," in which benefits professionals are asking how AI can improve every offering, rather than reaching for it as a bolt-on solution to a specific inefficiency.
The organizations moving fastest, he suggested, are those treating AI not as a replacement for human judgment but as a way to extend what advisers can do at scale. As large brokers acquire AI-native platforms rather than building in-house, the technology infrastructure question is increasingly being settled at the broker level, rather than the employer level.
On measurement, Owens's answer was deliberately low-tech: ask employees how they feel. Net promoter score and customer satisfaction ratings, he argued, capture something that utilization metrics miss - whether the tool is actually making a difference to the person using it.
"In the end, you want the employee to feel that it's actually bringing benefit to them," he said. "The best thing to do is you ask them."
