Case Study: Suppli

Suppli is rebuilding payments for the $500 billion construction materials industry. The Austin-based fintech gives independent distributors modern tools to manage credit, collect payments, and compete with national brands, powering billions of dollars in transactions since launching in 2021. They’ve stayed lean: one seed round, organic growth, and a remote engineering team spread across the U.S., Europe, and South America.

• Company: Suppli
• Founded: Austin, TX · 2021
• Team: ~17 builders across 3 continents

AI productivity, amplified

Suppli has been adopting AI tooling extremely quickly. As a small, lean team, they’ve been able to experiment with new workflows and bring on new tools much faster than most organizations. They’re now shipping 3x the pull requests they were before, and their hands-on CTO - David Paluy - expected the need for human interaction to go down. In practice, David’s observed the opposite.

The bottleneck for doing great work is no longer writing the code, but agreeing on what to build and how to steer the tools towards the right outcome. This requires more up-front alignment, and the team is using Tuple more than ever to get it.

“We create a very robust plan together,” David explains. “Then we let AI work on it, build all the stuff. When the result is ready, we run quality gates, check everything works as expected, and deploy to production same day.”

Tuple shows up most during the planning phase, with a few engineers on a call, talking through tradeoffs, feeding context into an LLM, reviewing the proposed approach, and tightening the plan before handing it off.

David also noticed the speed of learning increased using Tuple, which is easy to miss if you’re only focused on short-term output. When one engineer figures out a better way to use an agentic workflow, that advantage doesn’t help much if it stays in one person’s head.

“I pair way more on AI stuff with our team members. If some team member has a better idea of doing a workflow with AI, everybody wins. The speed of learning when you are just a single person is way less.” — David Paluy, former CTO

Pairing made the learning transferable, and Tuple made pairing easy enough that it actually happened.

The backstory

David joined Suppli as its founding CTO in 2021, after more than 25 years writing software. He stayed active in the codebase and with his team and was regularly pulled in when a production transaction didn’t look right.

But Suppli is remote, and engineers work across six U.S. states and multiple countries. As the team grew, David ran into a gap with his existing tools: the kind of quick, high-bandwidth collaboration he was used to in person just wasn’t happening. He couldn’t easily walk through a change with a junior engineer or pull two people in to debug an incident spontaneously.

They tried Slack first. It worked for coordination, but not for navigating code together. “I’d say ‘type this, no, not this, go there, not that one,’” David recalls. Audio reliability was also a problem. Sometimes he’d switch to WhatsApp just to get a call that didn’t cut out. Google Meet helped in some situations, but it still didn’t feel like sitting next to someone and working through a problem together.

The tools they were using weren’t failing dramatically; instead, they were creating just enough friction that folks wouldn’t collaborate unless it was absolutely essential and even then, it wasn’t enjoyable.

How Tuple became the default

David wanted something that made informal pairing feel normal again, like it is when working together in-person. David had heard of Tuple years earlier, so on a whim, he gave it a shot with a few engineers. Tuple quickly spread across the entire team.

“It’s a must-have tool for any engineering team that doesn’t sit in the same room together. You can’t really pair with anything else. Nothing provides the same quality.” — David Paluy, former CTO

Today, Tuple is the default for collaborative work. Whether the team is debugging incidents, reviewing pull requests, talking through architecture decisions, or walking through unfamiliar parts of the codebase, the team chooses Tuple over everything else.

The only real hurdle was getting people to stop reaching for the tool they were used to. Once the team had a few successful sessions with crisp audio and clear and snappy screen share, Tuple became the thing everyone reached for first.