AI in Benefits 101: One Year Later — What It Actually Takes for Employees to Trust AI with Their Benefits

In Part 1 of this series, we looked at what's changed in the AI landscape over the past year. In Part 2, we gave HR leaders a framework for evaluating vendor claims. This piece asks a different question, one that doesn't get nearly enough attention in the benefits AI conversation: what do your employees actually think about all of this?

It's also the question that determines whether an AI investment delivers on its promise or sits unused while employees route around it.

The Trust Gap Is Real, But It's Probably Not What You Think

Here's the framing most people reach for: employees are skeptical of AI, especially when it comes to their health. And there's data to support that instinct. According to Prudential's 2026 Benefits & Beyond study, just 24% of employees currently use AI for benefits guidance, despite its growing availability. Prudential's own researchers concluded this is "less of a technology adoption challenge and more of a trust challenge."

But here's where it gets interesting. The same research shows that employees aren't opposed to AI on principle. They're looking for tools they can actually rely on for important decisions. The behavior is already there as we see employees are using AI in their personal lives, for health research and for financial questions. What's lagging isn't willingness. It's confidence in the specific tools being put in front of them at work.

Justin Holland, HealthJoy's CEO, makes a similar point. He argues that the AI trust problem in benefits isn't really about AI at all. It's actually about wrong answers. "I don't believe a lot of people trust that when they call the carrier they're going to get the right answer and they're talking to humans there," he said. "Trust is inherent. If you believe something has great options and intelligence, you believe you're going to get the right answers."

In other words, trust follows accuracy. Not the other way around.

What Employees Are Actually Asking

The trust conversation becomes more concrete when you look at what members are actually bringing to AI. Across HealthJoy's platform, the most common member questions are about coverage, costs, networks, and finding care — the same questions that have always landed in HR inboxes, and the same ones employees have historically struggled to get fast, accurate answers to.

When an AI tool answers those questions correctly and quickly, trust builds. When it doesn't — when the answer is confident but wrong or generic when the member needs something specific — trust erodes. And in benefits, where the stakes of a wrong answer include unexpected bills and delayed care, that erosion is hard to reverse.

This is why the evaluation questions from Part 2 of this series matter so much at the member level too. An AI trained on generic public data can't give a member a reliable answer about their specific plan. An AI without a human backstop can't course-correct when it gets something wrong. And an AI optimized for someone else's interests like a carrier's or a PBM's, isn't going to earn the kind of trust that comes from consistently steering members toward what's actually best for them.

Trust Is Built in the Moment and Over Time

There's a parallel worth drawing here. Think about how trust in search engines developed. Early skepticism gave way to near-total reliance, not because someone explained how the algorithm worked, but because people kept getting to the right answer. The tool proved itself, interaction by interaction.

The same dynamic applies to AI in benefits. Employees don't need a detailed explanation of how the model works. They need it to work accurately, consistently, and in a way that clearly has their best interest at heart rather than someone else's.

That last part matters more than most vendors acknowledge. According to Prudential's research, employees are willing to engage with AI for benefits guidance provided they trust their employer to manage their data responsibly and that the tool is genuinely working for them. The moment an employee suspects the AI is steering them toward a more expensive option or giving them a confident answer that turns out to be wrong, the trust conversation becomes very hard to restart.

What It Actually Takes

Building employee trust in AI benefits tools isn't a communications challenge — it can't be solved with a better onboarding email or FAQ page. It's a product challenge. 

The tools that earn trust share a few things in common: they're accurate (not just most of the time, but reliably, on the questions that matter most) and they're transparent about their limits. When JOY doesn't have a confident answer, it says so rather than generating a response that sounds plausible. They have humans behind them, not as a fallback, but as a deliberate design choice that ensures the highest-stakes interactions get the expertise they require. And they're clearly working for the member, not for someone else's bottom line.

That's the standard we hold ourselves to at HealthJoy. And it's why we believe the path to employee trust in AI benefits isn't faster adoption or better marketing. It's better AI, built the right way, with the right people behind it.

Read more on why the best AI in benefits will always have humans behind it.