Marsh challenges the cognitive load assumption behind most workplace AI deployments
The assumption that AI reduces how much employees have to do is wrong — and the wellbeing programs sitting underneath most AI-enabled workforces were built for a different working environment.
That is the argument Kate Brown, Marsh's global digital leader for employee benefits, made on the company's Transforming Benefits with Technology panel series alongside chief technology officer Andrew Owens and head of product development Simon Jarvis. Brown's observation is direct: employees using AI are reporting that they are busier than before, not less so. The technology allows them to handle more work in parallel, increasing throughput rather than reducing overall load. The tasks do not disappear. The cognitive strain moves with them.
For HR leaders overseeing both AI deployment and employee wellbeing programs, that gap is the immediate question. A workforce that has adopted AI tools is not automatically less stressed or less at risk of burnout. In some cases it is more so, and the benefits package has not been revisited to reflect it.
Why the cognitive load assumption matters for HR
Brown's characterization — that some employers are still relying on mental health and wellbeing benefits created in 1984 — is a pointed way of framing a genuine structural problem. Employee assistance programs (EAPs), mental health benefits, and wellbeing tools were designed for workplaces where cognitive demands, pace, and the volume of parallel tasks looked materially different from what AI adoption is now producing.
HRD has reported on how too much AI is causing "brain fry" among employees — research suggesting adverse productivity outcomes once employees are managing more than three AI agents simultaneously, and that organizations which celebrate productivity gains without clarifying workload implications are amplifying stress rather than relieving it. Marsh's panel framing sits in the same territory.
The implication for HR is not that AI adoption should slow. It is that the wellbeing infrastructure sitting beneath it needs to be reviewed as part of the deployment decision, not after the fact.
How HR's role shifts as AI matures
Simon Jarvis framed the broader arc of AI adoption in benefits administration as moving through three stages: automation of discrete tasks, redesign of whole processes, and eventually the elimination of certain tasks entirely. At the far end of that arc, benefits decisions are increasingly made or suggested for employees based on data the organization already holds, with active employee choice progressively reduced.
HR's function shifts from running processes to governing whether AI is being deployed appropriately — a materially different role requiring different skills and a different relationship with benefits data. That governance question is already live: as HRD has reported, standalone tech that isn't integrated into core systems tends to generate administrative busywork rather than the efficiency gains employers expect.
Owens pointed to data connectivity as the near-term challenge. Most employers are running benefits across systems that do not share data effectively, which limits both the quality of AI-assisted decision-making and the ability to catch wellbeing trends before they become claims or absences.
The hyper-personalization opportunity — and why trust is the barrier
The panel identified hyper-personalization and digital twin technology as the developments most likely to reshape employee benefits over the next several years. A digital twin in this context is a model of an individual employee's likely future health trajectory, built from wearable data, biometric information, and lifestyle inputs, used to guide preventive benefit design rather than responding after conditions develop.
The opportunity is real. Preventive intervention at the right moment — before a condition becomes acute — reduces both the human and financial cost of employee health risk. For self-funded employers in particular, shifting spend from reactive claims to proactive prevention is one of the most commercially compelling arguments in benefits strategy.
The barrier is not technical. Employees will not share the data that makes personalization meaningful unless they trust the organization to protect it and can see clearly how sharing it improves their outcomes. That is a governance and communication problem, and most employers have not resolved it. Low participation rates render the personalization model ineffective regardless of the underlying technology.
What HR leaders need to ask now
Three questions follow from what Marsh's panel described.
First, has the wellbeing and mental health support in place been reviewed since AI tools were deployed across the workforce? The cognitive load profile of the workforce has changed. The benefits likely have not.
Second, do the benefits systems share data effectively enough to use AI-assisted decision support? That gap will widen as other employers invest in more connected infrastructure.
Third, for HR teams managing self-funded plans, is there an active conversation happening about shifting from reactive claims management to preventive intervention — and is that conversation happening before the next renewal locks in another year of the same program?
HRD has reported on research linking AI deployment to burnout risk among frontline workers, with 60 percent of employees believing AI adoption will increase stress and burnout, compared to only 37 percent of organizational leaders who see it as a concern. That perception gap is exactly the kind of problem Marsh's panel is describing — and it falls to HR to close it.
