Skip to content
Work Tech Weekly
Bjorn Reynolds

He Built an Employer of Record Company and Won't Let AI Make the Final Call

Bjorn Reynolds runs a company built on AI-powered HR tools. He's also the guy telling you not to trust them with the questions that matter most.

Bjorn is the founder and Chief Guardian (his actual title) of Safeguard Global, a global payroll and employer of record (EOR) company he built years before EOR became a thing. An employer of record doesn't just process payroll. It becomes the legal employer, compliance and liability included, in every country it operates. It’s also exactly the kind of business where "AI is usually right" doesn't cut it.

Make no mistake. He loves AI. He just doesn’t let it answer the employment law questions his own company is legally on the hook for.

Why AI Can't Replace Human Judgment in Employer of Record
  39 min
Why AI Can't Replace Human Judgment in Employer of Record
Work Tech Weekly
Play

How Safeguard Backed Into the Employer of Record Business

Before EOR was a category, Safeguard was just a global payroll company, running payroll in San Antonio with three employees and a founder who'd decided to stay "because someone's gotta deal with it." It didn’t set out to build an EOR business. It fell into one because it wanted to stop turning down money it didn't know how to accept.

Customers kept asking Safeguard to run payroll for workers in places like Brazil, except those workers weren't technically employees. They were contractors who wanted the same platform, the same experience, and Safeguard had no legal way to give it to them.

So Safeguard looked at the alternative — a PEO structure that barely existed as a concept in Europe at the time — and asked, “Hey, can we do this and keep the revenue?”

"We are losing all this business, and we're a payroll company," Bjorn says of the moment. "Can we do a PEO, so we still get the revenue and we'll just call it international PEO?" That's the whole origin story. The company got tired of saying no.

The origin story also explains what an employer of record is, structurally. Safeguard isn't consulting on employment law in Brazil or Namibia or wherever the client happens to need someone. It's assuming the actual legal employer role, in every country it operates, for every person on the books. And that's a liability, full stop.

Why Global Payroll and Employer of Record Are Different Businesses

After running a global payroll business and an EOR business side by side for years, the math eventually stopped mathing.

"We built the global payroll business, and we built an EOR business, two very different characteristics without a lot of crossover," Bjorn says. "The enterprise customers would buy EOR, but an EOR customer would never buy global payroll."

If you're a US-headquartered company with four people in Spain, you're an EOR customer through and through. If you need payroll run for 5,000 people, you may need global payroll, but you'll never be an EOR customer.

So Safeguard sold the payroll business to Deel in 2025 and kept EOR and core HR. It was a recognition that global payroll needed the kind of rollup-and-scale strategy Safeguard wasn't positioned to fund while also running an entirely separate business under the same roof.

Why does this matter for the AI conversation? Because payroll, as a category, never made you somebody's legal employer. EOR did. Liability doesn't live in the same place, which means tolerance for AI error doesn't either.

Being the Actual Employer Changes What You Can Let AI Touch

Even though companies across Work Tech are leading with the AI sizzle, the EOR companies still standing three years from now won't be the ones with the flashiest AI pitch. They'll be the ones that never lost track of which parts of the job humans need to answer for.

Safeguard ran a test that should make every compliance-heavy category nervous. It had large language models answering employment law questions, and they came back 86% accurate across 250 questions.

"You can be right 86 times out of a hundred," Bjorn says. "But those 14 is what loses you your job." For a company that's just software, a wrong answer is a bad review. For a company that's the legal employer of record, a wrong answer is Safeguard's own liability, not a client's.

Taking ownership of that liability is behind Safeguard's AI strategy, what Bjorn calls "human when it matters." AI can handle the benchmarking, the routine work of figuring out where a role should sit. But an employer of record agreed to hold someone else's liability. That's the whole model. So when a practitioner in the U.S. needs help with a dispute in Namibia, a person picks up. Not a bot. Because Safeguard is the one who'd actually answer for it if it went wrong.

Let's Talk