Henri Goes Outside: Our AI Agent is Turning Heads

Last week, we showed up to RacquetX (the largest racquet sports conference in the US) as a team of three: Russell, Julia, and Henri. Henri fit in Russell's pocket.

Henri is our AI agent with his own phone (he’s a loyal T-Mobile user with a real NYC-based phone number). When conversations got to the point where it was hard to explain in words what we were building, we just pulled out the phone and said “this is Henri.” 

That's when eyes started lighting up.

Henri is just one example of an agent we built specifically for padel coordination. You tell him you need a game (via text). He finds players in your network who match your level, proposes the options to you, waits for your approval, then reaches out to each of them by text – on your behalf – tracks responses, and confirms the roster. He texts like a person, organizes every conversation in a dashboard, and builds a social graph of your players over time so every match he coordinates makes the next one smarter.

The problem he's solving is real: getting four people in the same place at the same time, at the right level, with a court that's open, is genuinely hard. And every time a padel club has a court sitting empty, they’re losing money. 

One thing that kept coming up in conversations with clubs and platforms at RacquetX: app fatigue is real. People don’t want to download more apps, and clubs don’t want to have to onboard new ones, either. The platforms we spoke to want partners who can build on top of what they already have, not integrations that add complexity.

This is where agents change the equation in a way software alone can't. Henri doesn't require your players to download anything. He communicates through text, works within your existing tools, and connects to your data without forcing a new workflow on anyone. The nimbleness of an AI agent (the ability to plug into what's already there and just act) was one of the things that kept standing out in our conversations.

Padel requires a lot of coordination, so it’s a natural fit. But this pattern shows up everywhere.

We had a call this week with someone who advises businesses on growth strategy. He walked us through how almost any company's sales motion works: define your ideal customer, build criteria, do outreach. 

His example was one of his clients who sells golf simulators. The thesis was that country clubs in the Northeast were the perfect target – it's cold half the year and those golfers need their fix. But turning that insight into a list of actionable prospects means finding every relevant club, identifying the head pro at each one, tracking down their contact info, and starting from scratch if the first person doesn't respond. It's manual, it's repetitive, and for lack of a better term, it's soul-sucking.

An agent can literally do all of that. You still have to develop the thesis, but once you know who you're going after, the agent goes and finds them based on that criteria you’ve defined.

Henri runs on OpenClaw, the open-source framework that gives agents a persistent presence, a soul, real tools, and the ability to act autonomously. This week, Nvidia's Jensen Huang called OpenClaw "definitely the next ChatGPT." The distinction he drew: the shift from AI that answers questions to AI that autonomously performs tasks, makes decisions, and learns tools on its own is a foundational change in how work gets done.

Nvidia is backing that thesis hard enough to build NemoClaw, an enterprise-grade version with security, privacy, and oversight built in. When the company supplying chips for the entire AI industry also starts building enterprise infrastructure around a technology, that’s probably worth paying attention to.

Henri started as a solution to a very specific problem.

It turns out a version of that problem exists almost everywhere. Padel just happens to be where we let him loose first, but it won’t be the last place he fits.

Stay curious,

Julia & Russell