I spent six weeks building a presentation about AI.
I talk about summarising online classes, creating tasks in real time for educators, algorithms that tell a consultant whether a student is likely to renew before they pick up the phone, scheduling systems that predict what class someone will book next.
Today I delivered that presentation. In a conference room. In person. To franchise partners who flew in from Europe, Asia, Latin America, and the Middle East.
The most interesting thing that happened had nothing to do with the slides.
What the technology is actually for
There’s a ratio we track that I keep coming back to: the time a student spends studying alone versus time with teachers, coaches, and other students.
For many language learning products, this is 100%: students are always alone. Where teachers are available, they may spend at most 25% of their time with educators and other students. Our target is 50/50.
When I say that outside education circles, it surprises people. Shouldn’t the goal be self-sufficiency? More AI practice, less dependency on teachers? No. Students who spend more time with people learn faster, feel more supported, and engage at a higher rate. Language is a social skill — knowing every grammar rule doesn’t stop you freezing in a real conversation. Practice without another person in the room only gets you so far.
Every piece of technology I presented today exists to move that ratio. The platform makes it easier to book classes. AI tools free coaches from admin. The automated messages that fire before a coach gets involved are there to handle the routine so that when a coach gets in touch, it’s actually worth the student’s time.
That’s what the technology is for.
The conference as evidence
Between chapters — during coffee breaks and the Q&A pauses — the conversations had nothing to do with features.
A partner described how students may spend years at other providers trying to find the confidence to speak English in meetings. Two partners on opposite sides of the room discovered they’d solved the same problem in completely different ways and spent the break comparing notes.
I could have sent a recording. Everything I presented is documented. Partners from 32 countries didn’t fly to Barcelona to receive information. They came because information isn’t the point.
The point is what only happens in person: the conversation at the same lunch table that wasn’t scheduled, the moment someone’s body language tells you something their words won’t, the trust that doesn’t build from video calls alone.
This isn’t a case against remote work: I rely on distributed teams every day. It’s a case for being deliberate about what in-person time is actually for. When you bring people together from around the world, the scarce thing isn’t the data. It’s the attention.
Where AI deployments tend to go wrong
Two patterns I see again and again.
The first is automating the wrong thing. Removing friction for the customer can be brilliant. Removing the human from the moment where the human is the product is something else: cost-cutting dressed up as innovation. The question should be whether automating something creates more or less of what people actually came for.
The second is measuring the wrong thing. Engagement rates, task completion, time-on-platform: all useful signals, but not the one that matters. The one that matters is whether a person feels like someone understands what they need. You find that in a conversation, not a dashboard.
The AI deployments I’ve seen work consistently are the ones designed to free people up to have those conversations — not to replace them.
Two students, same starting point
I closed today with an image I’d started with six weeks ago, when I began building this presentation: two students, same motivation, same goals, same day one.
One has a platform that knows where they are and nudges them forward. A coach who sees them as a person. A centre worth making the trip for. A consultant who, when renewal comes, already knows how the last three months went.
The other doesn’t.
The difference between them isn’t ability or effort. It’s the system around them.
Everything I presented today — every feature, every workflow, every algorithm — exists to bridge that gap.
Leave a Reply