AI for Learning: Junto
How Junto combined a demand for innovation with a focus on outcomes to find new ways to improve their product.
👋 At our last online Edtech Fellowship meetup, we were lucky to host Allen Sanchez, Head of Learning Product for Junto who shared with us how they prototyped a AI Leadership Coach.
We’re running a meet up in London on Tues, 26 November. Sign up here. And offering 10% early bird discounts on our cohort programmes starting in the new year.
"It all started with a question," says Allen, Head of Learning Product at Junto. "What are you doing with AI?"
For Junto, an early-stage company specialising in cohort-based leadership training for new managers, innovation is often a necessity. In early 2024, clients began asking Junto’s founders, “What are you doing with AI?”
It was more than casual curiosity—leaders and CHROs across Europe were fielding pressure from their CEOs, eager to see how generative AI might make leadership development faster, smarter, and more impactful. Through light-weight prototyping, an idea emerged that resonated with prospective clients - an AI Leadership Coach for every manager.
Allen described the early days of exploration: “Our founders came to me and said, ‘Is this something we can build?’ As a product person who likes to lead with problems, not solutions, I instinctively pushed back. But we came to a place where we agreed: this was an opportunity where we should experiment.”
From idea to prototype in three days
In order to experiment quickly and effectively, Allen proposed to focus their efforts on the core product and integrating AI to improve the existing learner journey.
The goal was clear: design an AI leadership coach that could fit seamlessly into Junto’s established learning program—a series of biweekly, live sessions where small groups of early-career managers practiced the skills they’d need to lead teams effectively.
“This was always going to be scrappy,” Allen says reflecting on the need to move quickly. “We weren’t setting out to build a standalone AI product, but rather, to see how it could support our core offerings.”
Within three days, Junto had a working prototype, a Minimal Viable Test (MVT) powered by the ChatGPT API. The speed was intentional: “Move fast, but measure what matters” was their approach.
Bringing AI into the learner experience
The AI’s role wasn’t to teach. It was to provide a practice ground for leadership skills like communication.
“Role plays can be uncomfortable, nerve-wracking,” Allen says, acknowledging the tension learners often feel when rehearsing scenarios like giving critical feedback. With the AI prototype, Junto created a ‘safe space’ for practicing these challenging conversations.
In a typical session, Allen explained, learners would have time to practice the GROW model—a coaching approach for setting goals and exploring obstacles.
“Instead of jumping straight into peer role plays, we let people practice with the AI first. Six to eight minutes, and they’d get immediate feedback. Then they could start the real role plays with other learners with more confidence.”
The AI’s strengths aligned well with Junto’s needs: “AI can personalise very quickly to individual circumstances,” Allen said. “It gives people the chance to practice anytime, with guidance tailored specifically to the situation they’re dealing with. They can pick who they are talking to and add details about their work context to make it more realistic.”
Testing learner engagement with two key metrics
Junto needed to know if the experiment was working, so they chose two straightforward metrics: user enjoyment and perceived effectiveness. “We kept it simple,” Allen explained. “Were people actually enjoying this? And did they find it helpful?”
The responses were encouraging:
95%+ of learners enjoyed the AI experience, a strong indication that it was engaging.
90% found the feedback useful and relevant, validating the AI’s ability to deliver constructive insights.
By testing during the sessions in a supervised environment, they also gathered huge amounts of qualitative feedback which was encouraging and helpful to guide iterations.
Discovering unexpected value
As learners spent more time with the AI, in and out of the sessions, integrating it more and more in their daily routine, unexpected use cases emerged. “I thought they’d use it mostly for practicing leadership skills, but salespeople started using it to hone their pitches,” Allen shared.
Learners also found it valuable for preparing for difficult conversations—a testament to the AI’s adaptability. “Even I found myself using it to flesh out ideas before presenting to my own manager,” Allen admitted with a laugh.
One major surprise was the popularity of voice input. “We included voice input as a quick add-on, and it turned out to be the most popular way people interacted with the AI,” Allen noted. This insight, though unexpected, opened up new possibilities for future AI features.
Insights from rapid prototyping
Reflecting on the experiment, Allen shared several key lessons:
Stay focused on learner outcomes. “The mandate was to build an AI leadership coach, but we anchored everything around achieving specific learning goals,” he said. By keeping outcomes like engagement and behaviour change in focus, Junto avoided the pitfalls of tech for tech’s sake.
Listen to learners. “You have to trust your learners,” Allen said. Their candid feedback on what worked (and what didn’t) proved invaluable in shaping each iteration.
Don’t overthink—measure what matters. “We had dozens of metrics we could have tracked,” Allen says, “but in the end, we focused on engagement and utility. It kept us grounded.”
Lessons for the future of AI in learning
Allen’s journey with Junto’s AI coach offers valuable takeaways for others exploring AI in education:
Start with the basics. By adding AI to an existing program rather than building a new tool from scratch, Junto quickly identified what resonated with learners.
Experimentation fuels insights. Through rapid feedback loops and open discussions, the team learned to adjust the AI’s tone, approach, and even interaction style (such as adding voice input).
Integration over isolation. While standalone AI coaches are promising, Allen found that in this context they work best when supporting an existing learning journey. “Honestly, I don’t think standalone AI is the entire solution,” he concluded, reflecting on the experience. “It’s more effective when it complements real-life learning.”
The experiment may have been a Minimum Viable Test, but for Junto, the insights gained were anything but minimal. Allen’s final takeaway summed it up: “The results show that AI can support us, not as a substitute for human coaches but as a way to enhance the learner journey, making it more personalised, accessible, and impactful.”
Our next online meetup will be in January 2025 and Ian Ellard will talk to us about how the Opportunity Solution Trees and how helped the V&A Academy leverage learners insights. Sign up here if you wanna take part in the conversation!
Sounds like a really interesting talk! Developing something to aid real-world learning as part of an existing product (rather than building a standalone AI) + sticking to two key metrics are they key takaways for me.