How Gusto Built a New Product Line in 10 Weeks with Claude Code, No Jira, and No Docs
I sat down with Gusto's CTO Eddie Kim to learn how a tiny team of five shipped the new Gusto Co-founder product in just 10 weeks, throwing out all the standard product development rules and replacing them with a 24/7 Zoom room and AI.
Claire Vo
In the world of big tech, we’ve gotten used to quarterly planning, roadmaps that stretch out for years, and armies of people required to ship anything meaningful. So when I heard that Eddie Kim, the CTO and co-founder of Gusto, had spearheaded a project to build an entirely new product line in just 10 weeks with only five people, I knew I had to have him on the show.
I was so excited to sit down with Eddie because this story is a perfect example of a theme I'm seeing everywhere: executives are getting back to building. They're not just directing traffic; they're in the code, prototyping ideas, and fundamentally changing how their companies operate. Eddie’s experience building Gusto Co-founder is a playbook for how to do this. The team ditched nearly every tool and process we take for granted—no Jira, no Figma, no tech specs—and replaced it with raw building talent, a perma-Zoom room, and a whole lot of AI.
This wasn't just some side project. This was a tier-one launch for a company with over a thousand people in its R&D organization. Eddie’s story is about more than just speed; it’s about a new, more joyful and impactful way to build software. We walk through the radical development process they created, see the impressive AI agent they built for customers, and even watch Eddie use Claude Code to fix a bug live on the show.
Workflow 1: The 'No-Process' Process for Building an AI Product
The most fascinating part of this story is how the team built Gusto Co-founder. It all started, as many great ideas do, during an unexpected delay. Eddie was stuck with a five-hour layover in London and, instead of getting frustrated, he pulled out his laptop and started prototyping an idea that had been percolating in his mind using Claude Code. By the time he landed back in San Francisco, he had a working prototype.
He shared a Loom of the prototype with some senior engineers and a designer, Katie Chen, and the excitement was immediate. The team officially kicked things off during a company event in Denver with a single whiteboard session. That whiteboard photo became the only piece of documentation for the entire 10-week project.
Embracing the Void: No Jira, No Figma, No Docs
To move this fast, the team had to get rid of everything that wasn't essential. As Eddie put it, the process was more defined by what they didn't do.
“We had no meetings. We had no text specs. We had no Figma. We had no Jira board where we tracked stories or tracked work. We had no standups, no retros. We had nothing.”
Their entire operating system consisted of two things:
- A 24/7 'Perma-Zoom' Room: A constant, open video call where the remote team could pop in and out to collaborate, pair program, or just work quietly together. This replaced all scheduled meetings.
- Claude Code: The AI coding assistant was the engine that powered their development velocity.
The 'Trash Can Method' of Engineering
This new way of working also changed their relationship with code. The team adopted what I call the “trash can method,” where building a feature and then throwing it away isn't a waste of time, but a core part of the design process.
Instead of writing a PRD, an engineer would just build the feature and open a pull request. The PR itself became the proposal. The team would discuss it in the Perma-Zoom, and if it wasn’t the right direction, they’d just close the PR. Because Claude Code makes writing code so inexpensive, there was no sense of loss. They even threw out Eddie's initial prototype to rebuild the architecture from scratch using TypeScript and a Cloudflare Worker, which turned out to be the right call.
This process empowered everyone to build, including the designer, Katie. She started by shipping a “faked” front-end experience to production behind a feature flag. It had the full UI but returned canned responses. Then, engineers would progressively wire it up, breathing life into her design. Katie became such a prolific coder that she ranked in the 94th percentile for code throughput across Gusto's entire R&D organization. It's a testament to what's possible when you have a supportive team culture willing to mentor and review code from non-engineers.
Workflow 2: Automating Complex Payroll with Gusto Co-founder
So what did they actually build? Gusto Co-founder is an AI agent that helps business owners manage their HR and payroll tasks through simple conversation. A key insight was that many customers, especially small business owners, are always on the go. So, the team built it to be primarily accessible via SMS and Slack, not just a web app.
One of the most powerful examples Eddie showed me solves the problem of the “work before the work.” He shared the story of a massage spa in New York City. Every week, the owner had to:
- Export a report from their MINDBODY booking software.
- Manually calculate commissions and bonuses in a Google Sheet based on complex rules (e.g., $15 bonus for a 'Hot Stone' upsell, $20 for 'CBD oil').
- Pool and split tips among the therapists.
- Finally, enter all this data manually into Gusto to run payroll.
With Gusto Co-founder, this entire workflow is reduced to a single conversation. Eddie showed how he could just tell the agent the rules in plain English.
He gave the agent this prompt:
“Hey, Gusto Co-founder, I need you to run my payroll. Look at this spreadsheet that's called Export from mindbody. And here's how I calculate it. You know, for every hot stone upsell, add $15 a bonus for that therapist. For every CBD oil add $20 bonus for that therapist, we pull tips. So just take the group tips amount and divide it by, how many therapists we have.”
The agent then connected to the Google Sheet, performed all the calculations, updated the payroll amounts, and presented them for a final 'yes' from Eddie before submitting. It’s an incredible demonstration of how AI can automate tedious, error-prone tasks and give business owners their time back.
Workflow 3: Eddie's Personal Workflow for Fixing Bugs
Eddie wasn’t just leading from the sidelines; he was in the trenches coding alongside the team. He even walked me through his personal workflow for fixing a bug that came from customer feedback.
Test-Driven Development for AI
His process is a brilliant application of test-driven development (TDD) principles to AI. Here’s how it works:
- Start with the Problem: He identifies a bug, in this case, a confusing user interaction documented in a GitHub issue.
- Prompt with Context: He opens Claude Code in his terminal and, using a voice-to-text tool, dictates a detailed prompt.
His prompt is the key to the whole workflow:
“There's a customer issue that is outlined in this GitHub issue... Can you please read this issue and come up with a fix for the problem that's outlined here? I'd love for you to first write an eval that fails to show that you can reproduce this issue, then come up with a solution, and then prove that the solution works by showing that the eval now passes.”
- AI Writes the Test and the Fix: Claude Code reads the GitHub issue, writes a failing evaluation (a test case) that reproduces the bug, proposes a solution (usually a prompt change), and then runs the eval again to prove the fix works.
- Human Review: The most critical step is the last one. Eddie carefully reviews the code and the prompt change. He uses his judgment to ensure it's a high-quality, concise fix. Only then does he ask the AI to open a pull request.
This loop—human context, AI-driven test and fix, human review—allowed the team to move incredibly fast while maintaining quality. It's no wonder their median PR review time was just nine minutes.
The Future of Building
Eddie's experience is a powerful look at the future of software development. It's a future where small, empowered teams can achieve incredible things by leveraging AI as a true partner. It proves that by shedding old processes, we can not only move faster but also have more fun doing it. His advice to other leaders is clear: get your hands dirty. You can't understand how to build great AI products unless you're using AI to build them yourself.
I couldn't agree more. This is the moment for everyone to be a builder again. If you're interested in trying out what Eddie's team built, you can join the waitlist.
- Check out Gusto Co-founder
