The hardest question we ask before shipping anything AI-powered at Gusto isn't whether we can build it. It's whether a business owner trying to run payroll on a Thursday afternoon actually needs us to.
That's the bar. And it's higher than it sounds.
Every company has an AI story right now and most of them sound the same. We wanted Gusto's to be different, because if we get this wrong, someone doesn't get paid on time. Someone's health insurance doesn't go through. We don’t use AI because it's exciting, or because everyone else is, but because it’s the right solution for our customers.
So we asked the harder question: when does AI actually help the small businesses we serve, and when does it just add noise?
Here are the four principles that shape every AI decision we make.
Start with the problem, not the technology
The question we ask before building anything is: what problem are we solving for our customers? Sometimes AI is the right answer. Sometimes it's a process change that has nothing to do with technology. AI doesn't get to lead—our judgement about the customer problem does. If we can't articulate the specific thing someone is struggling with, and explain why AI addresses it better than anything else, we don't build it.
Keep businesses in control
Gusto is where people pay their employees, manage their benefits, and watch their cash flow. They need to stay in the driver's seat.
When Gus or Cofounder suggests an action, our customers decide, especially for anything consequential.
Transparency shouldn't require a deep-dive into our settings. The goal is to make AI's role visible in ways that feel natural: a label, a tooltip, a brief explanation when it matters. Trust is built through clarity.
A few questions we're actively working on:
How do we show what an AI agent is doing behind the scenes so people can trust it?
How do we design intelligent escalation, the moment something gets handed back to a human, without making it feel like a failure?
How do we preserve real agency when automation is making consequential decisions?
We don't have all of these figured out and our point of view on them will continue to evolve, alongside the needs of our customers.
Get better with every interaction
We pay attention to what's working. Direct signals (feedback prompts, ratings) and indirect ones (whether people actually use what we surface). We use that to make the AI more accurate and more useful over time.
But here's the part we feel strongly about: if someone takes the time to tell us something isn't working, that input has to go somewhere. Feedback collection that doesn't lead to visible improvement is worse than no feedback mechanism at all. It tells people their experience doesn't matter.
Design for everyone
A line cook getting paid on Friday and a CFO running a 200-person operation both need this to work. So does the nonprofit director hiring her first employee. So does the plumber paying her crew.
Our AI has to work fairly for all of them. That means training on diverse data, testing for bias, and building real channels for customers to flag when something feels off.
It also means our AI outputs need to be written clearly with no jargon, no assumptions, nothing harder to understand than the original problem. Accessibility shapes every decision.Fairness in AI is ongoing work. We're committed to it.
Why we care about getting this right
Most Gusto customers are business owners trying to make payroll on time, stay compliant with rules that keep changing, and take care of a team they genuinely care about. They don't have margin for tools that add confusion or erode trust, even unintentionally.
That's the context we carry into every AI decision. Every call starts in the same place: what does this person actually need, and are we confident this helps?
We don't always get it right on the first try. Gus and Cofounder are still learning what to surface and when to step back. We're still learning where AI helps and where it just adds noise. But every decision starts with the person on the other side of the screen, trying to take care of their team.
That's the part we don't want to get wrong or ever forget.
