Steve Jobs once called the computer “a bicycle for our minds”—a tool that multiplies what a person can do. AI is a major upgrade to the idea. The bicycle is now a rocket.

Obviously, a rocket beats a bike, but the trick is knowing how and where to use that rocket.

At a panel on AI at bswift’s Idea Exchange conference in Nashville, a group of experts discussed just that. Chun Schiros, a field CTO at Amazon Web Services who advises companies on AI governance, distilled the mistake many are making into one vivid image. “You’re bolting a rocket engine on a bicycle,” she said. Yes, you move faster, but you’re spending a fortune to do it, and you’re still riding down a bike path.

So, she swapped vehicles. “Think of AI as a monster truck,” she said. “It has these really intelligent engines, which is the large language models, and it can run any errands that you throw at it.”

But AI needs the right platform to reach its potential—a highway, not a bike path.

“The highway is not going to just take you from A to B faster, but take you to C, D, E, F, G, places you cannot imagine before,” she concluded.

Why AI in Benefits Administration Is Finally Ready for Prime Time

AI is grabbing headlines and getting executives to sit up in their seats, but the term isn’t new—it dates to the 1950s. But only recently have three factors converged to make it possible:

• Massive amounts of digitized data from the internet era gave machines something to learn from.
• The math caught up, enabling breakthroughs in machine learning architecture.
• Cloud computing made training and running these models economically feasible.

Brad Flippo, a consultant with herronpalmer, summed up agentic AI as “a Ph.D.-level intelligence with a fraction-of-a-penny cost, and you don’t need to train it for 12 years.”

“It is getting more and more capable, and it’s getting cheaper and cheaper, which makes these ideas of, ‘Hey, let’s leverage it to create something useful,’ become economically viable,” Schiros said, tying it to the bottom line.

Schiros was blunt about the cost of waiting. “Speed is the ultimate business advantage,” she said. “The cost of waiting is the relevancy of your business.” But she added a caveat in the same breath. “The cost of running blind is also really high.”

Stop Asking If They “Have AI,” Ask About AI Governance in HR

That’s why AI governance is an essential buying question. HR needs to uncover what governs the AI, and who owns the quality of what it produces—because the honest answer still includes people.

The fastest way to get fooled by a slick benefits admin platform demo is to ask, ‘Do you have AI in there?’ Everybody does, or at least says they do.

Better questions to ask when evaluating AI in benefits administration are:

• What measurable change does this create in my workflow?
• How does it improve the employee experience?
• What quantifiable business outcome does it produce?
• What governance and guardrails ensure it’s safe, explainable, and reliable?

If a vendor can’t answer those questions, they’re selling you the rocket engine and hoping you don’t look down at the bike.

What AI-Native Benefits Administration Looks Like in Practice

Standing up an HSA plan is the kind of project that gives benefits teams migraines. The old way is weeks of work—defining requirements, building the carrier and payroll feeds, and testing eligibility scenarios one at a time—and hoping nothing slips through. On a platform built for AI from day one, that work runs differently.

“AI really shows up in the work, and not the way that things are worded,” Flippo said. The system drafts the configuration from your plan documents, maps the file feeds, and runs test cases—and then a human validates the outputs. It’s also proactive in catching things like payroll feeds that don’t reconcile, flagging them in the moment.

That precision matters because benefits administration is high stakes. A missed detail isn’t an asterisk, it’s somebody’s coverage. And when HR teams are freed from that kind of work, an individual contributor becomes, as Tanner Pratt, bswift’s SVP of Data and AI and the panel’s host put it, “a super contributor.” They get hours back for the strategic work that moves an organization forward.

From Benefits Access to Benefits Activation

On the employees’ side of the equation, an AI-native benefits platform goes after a different kind of waste—the benefits you’re investing in that employees aren’t using. Companies invest a lot, but they aren’t seeing the ROI in their employee benefits packages.

Schiros took a moment to illustrate the point, sharing how she only discovered a gym membership benefit she had through work the day she was leaving the company for another opportunity. AI can help close that gap between the benefits an organization invests in, and the benefits employees find, understand, and use.

“And it’s not about ‘let’s do another training about the benefits we offer,’ it’s about putting those benefits into action,” she said.

That’s the idea behind bswift’s personalization engine, Evive™, that nudges employees in the moments that matter, surfacing relevant benefits—helping activate a company’s benefits investment so the gym membership gets found while it still matters. It’s also the goal of Emma™, an AI-powered benefits decision support tool that can guide informed enrollment decisions and answer questions about an employee’s benefits package.

What HR Leaders Should Do Next

The starting point with AI is making sure it has an open road. Look at the platform, the workflows, and your goals—the things you’re trying to make easier.

“This is not the sales pitch of like, ‘Woo-hoo, AI is the best thing ever,’” Pratt said as he closed the panel out. “We want to help you innovate within your organizations and deploy AI the right way so that you have the impact that you can have thoughtfully.”

Get the road right, and the rocket (or monster truck) will take you places the bike never could.