Everyone wants personalised learning. Almost nobody is talking about the data.

Everyone is selling personalised learning. It is the oldest dream in education: teach a young person as who they actually are, instead of as the middle of the class.

I have been building in this space for over a decade, and the same thing keeps striking me. Almost nobody selling personalisation talks about what personalisation actually runs on. It is not a model. It is not a chatbot. It is data. And the closer I get to educational institutions, the more I find the data is a mess.

What the mess actually looks like

Picture a student asking a simple question: what are my career prospects?

One authority looks only at their exam results and points them down a road. Another looks at their holistic profile and tells them to go and explore. Both answers sound authoritative. Neither is derived from actual recruitment data. Neither checks where the money is genuinely being invested, in corporates or in startups, to see whether the advice survives contact with the market. So a young person walks away with confident direction that has no connection to the economy they are about to enter.

The second half of the mess costs just as much. Because the data is scattered, that student cannot put a structured portfolio in front of anyone. Everything they have ever done sits in a different system, so an employer has to reconstruct them from twenty five or thirty documents. It could be one place: organised, illustrated, understood in a minute. Instead everybody loses.

That is the real state of personalisation. It is not a technology gap. It is a plumbing gap.

The mapping framework

What has to exist underneath all of this is a mapping framework: a way to connect what a student actually does to what it says about their skills, and to where that leads.

This is not a Hong Kong story. I have been working with institutes across Europe, England and Southeast Asia, and having the same conversations at home with bodies like the Examinations and Assessment Authority, the Jockey Club, and the nonprofits pushing in the same direction. Different countries, same gap. We have also gone deep on the standards published globally.

They share one flaw. They map from a human perspective. They were written for a human to read, interpret and apply. But the world they are being deployed into is automated, which means they need far more technological readiness than they have. An AI has to be able to ingest and apply the map at scale, rather than follow rules that someone with a title set once and never had to defend again.

They also cannot be written once. The rules need to be cross referenced and scrutinised constantly so they move as the world moves. This is where academia struggles, because most of the research publishing these frameworks was never built to be continuously challenged and updated. Industry is. The market corrects you every quarter whether you like it or not.

Then there is the layer past machine readable, and I think it is the one that matters most. The map should not be a flat set of rules saying this activity equals that skill. It should be structured and visualised so that what you are mapping is the correlation itself, along with the explanation behind it. Why does this connect to that?

That distinction decides what a model can do with it. Hand an AI a rule based set and it spends its time executing someone else’s logic, then stops. Let it see why a mapping was made in the first place and it can interrogate that reasoning, test it against what is actually happening in the market, and evolve the map on its own.

So the ask is not a better taxonomy. It is a machine readable map of correlations and the reasons behind them, wired to real recruitment data, scrutinised like a market instead of published like a paper. Build it that way and it does not just get applied at scale. It gets better without you.

Why nobody has solved it

The reason is not technical, and I would rather say it plainly.

Education is a slow and painful industry to move. It is thick with politics, egos, and in places, cliques that do not let new people in, so good ideas lose to whoever the person handing out the opportunity already knows.

It is also too easy to say education is behind. Plenty of people in the working world are just as behind. This is not an education problem. It is a change problem.

And realistically, this is not a ten or twenty or thirty year fix. It might take half a century. What that asks of the investment side is people who understand distribution strategically, and who understand the scale you have to reach before the impact even becomes visible. Those are the people we work most closely with. At BSD, building the foundation took us over a decade. Only now are we slowly turning the knob that turns it into impact.

Meanwhile, young people have no direction

While the plumbing gets built, there is a live problem in front of us. Teaching skills is no longer enough, and this is not a Hong Kong story or a US one.

And here is the number that should stop anyone selling education. In low and lower-middle income countries, young people with advanced education are far more likely to be unemployed than those with only basic education. Nearly four times more likely in low income countries: twenty one percent against under six. More education, less work.

That is not a story about lazy graduates. It is a story about a map that never connected what people learn to where the work actually is.

The pattern I keep seeing up close is the sharpest version of it. The most studious graduates, the ones who protected their GPA, are often the ones who skipped the internships entirely. They optimised for the metric the market was quietly retiring, and walked into a commercial world they had never once touched.

What internships gave was direction, not a line on a CV.

They put you in a room with people you did not choose, on a problem you did not set. They put you near work with money riding on it, which is the thing a classroom cannot simulate. In school, your work is judged on whether it is correct. At work, it is judged on whether it was worth doing: what it cost, what it returned, and why you chose it over the three other things you could have done with the same week. That is a different muscle, and it only grows under real conditions.

Most of all, they gave confidence. You watched something you built leave your hands and get used by real people. For me that is the single biggest difference between the work you do in school and the work you do in the real world, and no amount of coursework replaces it.

That is what a guidance counsellor spent the last twenty years trying to deliver, with a caseload of hundreds and almost no data to work from. Which is why there is a need for something like a coaching tool, the kind of thing I am actively working on, to amplify that person rather than replace them.

What personalisation actually means

Personalisation is not a prettier dashboard or a friendlier chatbot. It is a map: from what a young person does, to what that says about them, to where the world is actually putting its money.

Build the map machine readable. Wire it to real recruitment data. Make it explain itself, so it improves without waiting for a committee. And back it with people who can sit through the scale before the impact.

In the meantime, if you are a student reading this: go and get the experience. Your GPA was never going to hand you your confidence.