Your Employees Don't Trust the AI You Just Bought

Sixty-seven percent of employers want to use AI to help workers make better benefits decisions. Only 28% of workers would actually trust what it tells them.

That's not a technology problem. That's a design problem.

Forrester calls it the "employer-first, employee-lagging" pattern. Employers buy AI to cut call center volume and speed up enrollment but workers want something that feels personal and has a human backstop. Both things can be true but right now, most deployments are built for the first group and handed to the second.

Here's where to start closing that gap.

First, look at what your AI tool actually says to employees. Not the pitch deck version...The real output a worker sees when they ask about their deductible or whether they should pick the HDHP. Is the recommendation just an answer, or does it show its work? Benefitfocus found that employees who received a clear rationale for AI-generated suggestions were twice as likely to act on them. Accuracy isn't enough and people need to understand why.

Second, take the privacy concern seriously instead of dismissing it. Fifty-eight percent of employees worry about data privacy when using AI-assisted benefits platforms, according to MetLife's 2024 Benefit Trends Study. That's not irrational fear. These are people sharing health status, dependent information, financial decisions. If your rollout materials don't address what data is used, how it's stored and who can see it, you've left a real question unanswered.

Third, keep humans in the picture. The 28% trust figure doesn't mean workers hate AI, it means they don't want AI instead of a person. Position the tool as a first stop, not a final answer and make it easy to escalate. "Talk to a benefits advisor" should be one click away, not buried in a footer.

Fourth, pilot with a group that will give you honest feedback...Not your most tech-comfortable employees and not HR. Find the people who usually call the benefits line with questions and watch how they interact with the tool. What confuses them? What do they skip? That's your real UX data.

Fifth, close the loop after open enrollment. Did employees who used the AI tool make different elections? Did they call in less? Did anyone complain that the recommendation felt wrong? You need that feedback to know whether the tool is actually helping or just running in the background while everyone ignores it.

My shovel metaphor holds strong here. AI can do a lot of the heavy lifting in benefits support, but only if it's built for the person swinging it...not just the person who bought it.

The employers who get this right won't just reduce call volume. They'll actually help people make better decisions about their health coverage. That's the whole point.

If your benefits AI is live but adoption is flat, it's worth asking: do your employees understand what it's telling them, and do they believe it has their interests in mind?