The AI Guild: What We've Learned Aligning AI Fluency with Judo Belts

Three months ago, 208 Rydoo employees sat down and rated their own AI maturity, from white belt to black. It wasn't the first sign of appetite for this: 77% of us were already experimenting with AI individually before the system existed. It was the first time we measured it, and turned enthusiasm into something we could actually track and act on.

Building an AI-first culture at Rydoo

Sebastien Marchon, CEO of Rydoo, wrote about how we built an AI-first culture at Rydoo: the Guild, the tool stack, the weekly rituals. With insights from Esteban Arias, Head of Product Marketing, and Kelly Klein, Senior Sales Enablement Manager, who were both pivotal in the development of the AI Guild and subsequent initiatives, this piece goes a level deeper, into the belt system itself, through the eyes of the people closest to it.

Why we chose a judo belt system to align AI fluency

Sebastien has described the origin plainly. Earlier this year, Leadership set out to make sure AI wasn't just a product and engineering tool at Rydoo, and Sebastien kicked it off by asking a specific question about himself: what was his own AI maturity, next to Fernando's, next to Daniela's, next to anyone’s at Rydoo? That's why the system needs to measure individuals rather than departments. "AI is now part of the normal skills," Sebastien remarked at Forge Connect in Oslo earlier this year, "there is no success if there is no measurement."

The company surveyed employees first, then built the AI Guild, then commissioned the belt system to answer the maturity question directly. Of 255 employees invited to self-assess, 208 completed it: resulting in 15 white belts, 68 yellow, 76 green, and a smaller group across blue, brown and black. Sebastien is upfront that this isn't a report card: "The goal is not to shame people. Being a white belt is fine. What is not fine is being a white belt at the end of the year if no learning has happened."

"The belt system works because it's a familiar developmental construct that is accountability-based," says Kelly Klein, Senior Sales Enablement Manager, who took on building the self-assessment itself. Competencies by belt are meant to be clear, and the decision to attempt the next level sits with the person, not a manager. The hard part is calibration: if you are a green belt yourself, how do you define what a brown belt should look like?

"I would never have imagined myself coding, vibe or otherwise," Kelly says of building the assessment. What came out of it is self-reported: employees read scenarios about how they'd behave with AI and pick the closest match, with higher belts requiring a written explanation. It would be possible to skew the results, but not worth it, since a brown or black belt is expected to produce work, complex workflows, agentic support, vibe-coded tools, that a novice can't fake under scrutiny.

By Kelly's account, the assessment did three things well:

• it got the whole company thinking about how it uses AI, not just whether it does;
• it gave Rydoo a real baseline instead of a guess;
• and it means the AI leadership Rydoo is now hiring inherits a clear starting picture rather than a blank page.

The AI Guild that stewards all this is deliberately small and rotating: 12 employees, drawn from every function, from GTM to legal to finance, and nominated by department heads rather than picked for existing AI skill, with membership turning over regularly to keep decisions from happening in a vacuum.

How the AI Guild changed things from a sales perspective

In this section, Kelly Klein, Senior Sales Enablement Manager, shares her first-person perspective on the AI Guild's influence:

“I lead Sales Enablement, with a background in learning and development, and I came into this as someone still building my own AI fluency rather than someone who already had it. 

That's part of why I gravitated toward the AI Guild's people side. I'd watched AI change how I worked myself, and wanted to help others take the first step.”

“We landed on the name TGAIF, Thank God It's AI Friday, for our bi-weekly whole company AI sessions which involved AI workshops, departmental use case demos, and more. We also have multiple AI channels on Slack for sharing tips. What's changed since the development of the AI Guild, plus TGAIF, and these channels, is that individual teams now run their own sessions and swap tips across channels without anyone asking them to.”

“The clearest before-and-after I can point to in sales is MEDDICC, the deal diagnostic framework I was brought in to operationalise across sales. We were barely using it when I joined. We built an AI Skill into our call recording and analysis tool, paired it with coaching, and reached almost 100% adoption, quickly and efficiently. It is a game changer in forecast accuracy and pipeline health.”

How the AI Guild changed things from a marketing perspective

In this section, Esteban Arias, Head of Product Marketing, shares his first-person perspective on the AI Guild's influence:

“I sit close to the AI Guild's day-to-day, and the biggest surprise wasn't a tool. It was the makeup of the group. The instinct is to staff a guild like this with your AI experts. We didn't. Mixing proficiency levels and functions eliminated a lot of assumptions that would otherwise have shaped a one-size-fits-all rollout. The question changed from "how do we get people to use AI" to "how do we help them be effective in their own context."

"While running the first sessions we understood that this was a change management transformation rather than a technology one."

“The pace has been the real surprise. Someone using AI in the most basic way one week is often building automations and agents the next. The AI Guild was never trying to make everyone a black belt. The goal was proficiency that produces results.”

How different functions are influenced by AI fluency

Marketing used AI to run more personalised campaigns and free sales up to be self-sufficient. Product used it to prioritise backlogs and turn ideas into working prototypes faster. Sales, per Kelly's MEDDICC example, used it to shorten learning cycles and surface deal risk earlier. Operations used it to speed up onboarding and recruitment admin. Legal used it to review documents and move contracting along faster.

Finance's shift is more definitional than anecdotal so far. Sebastien has described a white belt in finance as someone using AI mainly as a convenience tool, and a brown or black belt as operating at optimisation level across the wider finance function, sharing knowledge and educating the team as they go. By his own account, finance sits slightly behind product development in AI maturity today, which he calls unsurprising rather than concerning.

Esteban noticed the same pattern across the functions that have moved further: creativity first, then collaboration. People tried new approaches before anyone told them to, then shared what worked. Product started working with CS. Marketing started working with Operations. Nobody had to organise that.

What a black belt in AI fluency actually looks like

Nobody's reached one yet. Kelly's own definition is expansive: "An AI Sales Black Belt uses AI not just to optimise but to transform sales. They build AI-native workflows, create systems that scale across teams, and shift human effort toward the work that matters most: strategy, judgment, and customer relationships." Most importantly, she adds, "a Black Belt works to elevate their team and company's AI usage by sharing their knowledge." As she puts it: "I don't think we have a confirmed black belt yet at Rydoo, but because of the belt system we are all much more aware of what we individually need to do to get to the next level."

Esteban's read is similar: "the difference is that a black belt will understand if the tool should be built in the first place, if it will solve a problem effectively or generate usage without real impact." Most people can already get AI to do something. Deciding whether it should do that thing is the rarer skill.

What's next for Rydoo's AI-first Journey

Kelly wouldn't change much about the approach itself: waiting for perfect conditions would have meant less struggle, however, it would also have meant much less learning too. What she's watching now is quality control. As the company keeps creating and sharing prompts and skills, her own question is what happens if some of what's being trusted turns out to be out of date or simply wrong, and how you keep that quality up while still moving fast.

For now, Rydoo is erring on the side of less caution, which she calls "a good way" to create an environment ripe for exponential growth. In her words: "With great (and fast) growth comes great responsibility, but I'm sure we'll manage. The hardest part was taking the first step."

The next stretch in Rydoo’s AI-first movement is less about one system and more about four moves happening together:

• spreading AI knowledge further through the Guild, champions and a growing internal use-case library;
• encouraging more experimentation, building on the hackathon habit and the weekly rhythm TGAIF already keeps alive;
• embedding AI more directly into daily tools rather than treating it as a side project;
• and backing all of it with real resourcing and leadership commitment, not just intent.

A dedicated AI leadership function is part of that next step too, brought in to build on the baseline the belts already gave the company, rather than start from nothing.

The result of Rydoo’s efforts over the past year were demonstrated live in July. when six Rydoo colleagues spent 24 hours at Google's Paris offices for the Eurazeo AI Hackathon. In 24 hours, they built ‘Gotcha!’, a live assistant that answers customer questions mid-call, and they walked away with the Grand Prix against 16 other companies with a business case that uses AI to save more than €300,000 a year. True ROI that not many companies can boast despite their AI efforts.

The belt system and this AI hackathon example are the same bet: give people room to try, then judge the results.