Breaking down the complexities of global payroll is, to put it mildly, not for everyone. It’s like durian. Or licorice. Or heavily peated Islay scotch. Or Cookie Monster death metal. It’s an acquired taste.
You may be saying, OK, you do you, but … what do you think is so interesting? Maybe it’s that the stakes of payroll are getting it right every time, or you’re in deep … trouble. Maybe it’s that it is the highest level of complexity in HR/Work Tech. Global payroll is like herding cats… except every cat is a different country, your Core HR system is the litter box, and every cat keeps changing its preference for kitty litter. Oh, and maybe it’s that building a global payroll engine is the Maltese Falcon for many a tech founder — the stuff dreams are made of.
On this episode of Work Tech Weekly, we’re tackling the complexity of the global payroll challenge with Jerome Gouvernel, CEO and co-founder of datascalehr. Jerome is one of the smartest guys I’ve ever talked to about global payroll. He’s not some hoodie-wearing, hubris-infused tech bro. Jerome helped build ADP's GlobalView platform 25 years ago, one of the first global payroll engines across 45 countries. He’s legit and able to cloud the situation with some facts.
And Jerome has a perspective. He spent 20 years building and refining global payroll systems before he realized the industry had been solving the wrong problem. He has experience … and some thoughts.
It’s like this: Imagine that your global payroll implementation was supposed to take 12 months to complete and is now going on Year Two. The vendor says it’s a data quality issue. Your team says the vendor’s platform can’t handle the complexity.
They’re both wrong.
The maintenance burden of true compliance across dozens of countries and different ERP and HCM providers is simply unsustainable. No vendor gets the volume to justify thousands of annual rule changes per country. The knowledge required for each implementation with core systems can’t be reused. Yet the industry keeps selling it anyway.
The breakthrough came when he stopped thinking about integration as a language-learning exercise and started rethinking the entire approach. “What you’re trying to do is not learn a new language,” Jerome says. “What you're trying to do is transact.”
Jerome joins the show to explain why this pattern persists and what it would actually take to break it. We cover the insanity of the current approach, why “full compliance” promises are economically fragile, what enterprise AI actually requires to be trustworthy and how there’s a better way to do global payroll.
Simply put, traditional payroll implementations treat every connection as a custom translation project. Imagine you are trying to connect your Workday HR system to your French payroll system. Teams spend months learning the vocabulary, field structures and business rules of both systems just to move data once. Next time, when you connect Workday to a Tanzanian payroll provider, you start from scratch.
“At least when you're learning French, it might take you two years to learn it, but then you can speak to every French person in the world,” says Jerome. “In the world of ERP, you're learning that language. It takes the same amount of time, but next time you need to do it, you're going to have to learn another brand new language.”
This is why implementations are still waterfall like it’s 2006. Teams spend eight months on data preparation just to run the first parallel payroll test. That fails. You have a bunch of panicked meetings. More prep. Repeat until morale dies. You can't iterate quickly when it takes half a year just to move data from one system to another.
The challenge, as Jerome describes it, isn’t a systems problem. It’s a reality problem. Take a U.S. zip code. It doesn't exist in a French payroll system. The traditional approach is to learn what the French equivalent concept is, what it's called, and where it lives in the data structure. Jerome spent years refining this process, and the conclusion was always the same. “It never felt elegant,” he says. “It just felt like this is a brutal approach.” And it was. The knowledge is specific to one vendor, one client, one country. It's not reusable. Every new connection is a very expensive, very frustrating little snowflake.
The non-reusable knowledge problem is bad enough. But it gets worse when you look at what vendors are actually promising to maintain.
Jerome’s work on ADP's GlobalView platform proved the technology could work. But then reality reality shows up with a lawn chair, an attitude, and a cooler full of cheap beer. You're not just maintaining 45 payroll engines. You're maintaining a promise of full compliance. For all employment types. And all industries. And all edge cases. Forever. But wait. There’s more.
Every year, governments make tens of thousands of trivial changes. A tax form moves a field. A new subtotal appears. Someone has to notice, understand where the data comes from, build the new form, and maintain version control. “None of it is particularly difficult to do, but this is just one form in one country, and then you just keep adding that,” Jerome bluntly says.
No vendor gets the volume to justify that effort. But, here’s the punchline: They keep selling it anyway.
So, now there’s a new shiny object. GenAI will solve the maintenance problem. It will simply read government websites, extract the rules, and write compliant code automatically. Cool. Love all that “self-driving car” energy. So, problem solved?
Jerome says hard pass. “If that's your idea of compliance, you're going to be in serious trouble,” he declares. You can ask Claude to simulate a German payslip right now. It takes 90 seconds and might even look convincing. “Is it accurate? You don't know. Would it ever work in real life? No.”
Not that AI isn’t going to play a big part in the future. It’s just that you can’t use it like an 8th grader ChatGPTing a science fair experiment real quick so he can get back to playing Fortnite. If enterprise software is going to use AI meaningfully, it needs infrastructure that traditional ERP architectures weren't built to provide: the granular auditability of AI decisions.
Auditability requires a fundamentally white box framework, according to Jerome. “Every decision along the way that AI participates in has to be explainable, auditable.” It’s acceptable to be wrong as long as you can trace where it broke and fix it.
The problem is that traditional data models don't offer that granularity. Why was a pay code mapped this way? Who approved it? What happened next? That chain of knowledge has no natural home in traditional database infrastructure. So, before you can have an AI strategy, Jerome argues, you need a data strategy. That’s where things get interesting.
Way too much AI buzz is about magic prompts, vibe coding, and front-end heroics. Real AI transformation is happening at the data strategy level. This insight unlocked datascalehr's approach. Instead of forcing teams to learn entire system languages, they break the problem down to field-by-field mapping. The result? A payroll admin can build a connector between Workday and a Tanzanian payroll system in two hours because they're not learning a language — they’re mapping individual fields. The system remembers those mappings and makes them auditable.
That granularity enables iteration. Instead of eight months preparing for a first test, teams can move data in an hour, run payroll, analyze gaps, and repeat the next day. Within five or six iterations, Jerome says, you're already at 99.9% accuracy. That’s how modern systems should work.
So, Jerome’s story is basically, “I spent two decades optimizing the wrong thing … until I realized the question itself is wrong.” Global payroll doesn’t fail because clients are dumb, vendors are evil, or the work is too hard. This stuff fails because the model itself is structurally unstable.
Solving this problem — or any enterprise software problem in the age of AI — starts with questioning our assumptions. That’s part of why I like talking to Jerome. I’m glad to know there are people out there questioning everything in pursuit of new solutions. That’s the true spirit of innovation. I’ll raise a glass of Lagavulin to that. But you do you.