Digital Transformation in EdTech 

Three years ago feels like a different geological era. That is not hyperbole about technology in general, it is specifically what AI has done to education. It has changed how learners study, how providers teach, and how those of us running online services acquire and keep an audience. Anyone in further education, apprenticeships or skills delivery who is still treating this as just an IT upgrade has already fallen behind.

In my experience, the biggest challenge isn’t underinvestment. It’s seeing organisations run digital transformation as two disconnected programmes: a shiny learner-experience project on one side, a back-office efficiency project on the other. They are the same system. Every hour an admissions team spends chasing an unqualified enquiry is an hour not spent supporting a learner who was ready to enrol. Every clunky internal process eventually surfaces as friction in someone’s learning journey.

Here are five areas where that connection plays out in practice, with what we have learned building them at Docsity.com, a study platform used by millions of students, and WeUni.com, which connects learners to institutions.

1. Hyper-personalisation: ask for one thing, at the right moment

Static, one-size-fits-all pathways are finished. But most organisations pursue personalisation backwards, they build a long registration form, ask for everything upfront, and watch completion rates collapse.

We took the opposite approach with a progressive profiling project on Docsity.com and WeUni.com. Rather than one big form, we identified the moments where a user is already engaged and where a single question or piece of data is genuinely useful to them: during brief wait times as they use the service. At those moments, we ask the single thing that matters most to both of us. Users answer, because the answer visibly improves their results within seconds.

Personalisation data should be collected at the point where it immediately pays the user back. If you cannot explain what the learner gets in return for the answer, do not ask the question.

2. Omnichannel isn’t one message on five channels

Meeting learners where they are means accepting that the same person behaves like two different users depending on the device in their hand.

When at Docsity.com we analysed the usage of our study tools by channel, the split was stark. Mobile app usage skewed heavily toward capturing something live. Web usage skewed toward filling content gaps or using structured study tools. Same brand, same learner, completely different jobs to be done. Had we designed one experience and pushed it everywhere, we would have served both badly.

Take WeUni.com, for instance: on a mobile device, users might voice-chat or experiment with the app, whereas during desktop sessions, they are fully focused on the task at hand.

Segment by observed behaviour and context, not just by age bracket. “Gen Z prefers video” is not a strategy. “Our app users are capturing; our web users are revising” is.

3. Automate the filtering, not just the sending

The automation conversation in education usually stops at chatbots and email sequences. The higher-value target is the volume of low-quality work that generic marketing generates in the first place.

Many institutions we work with are not short of enquiries, they are drowning in them. Broad campaigns capture people with no real intent, admissions teams burn weeks qualifying them, and budget disappears into clicks that never convert. Our approach with WeUni.com is to intercept learners at the exact moment they are studying and actively researching their next step, so the enquiry arrives already contextualised. The automation that matters is not the follow-up email; it is the qualification that happens before a human is ever involved.

Before automating a process, ask whether the process should exist at scale at all. Automating a bad funnel just produces bad outcomes faster.

4. Iteration speed is a cultural problem before it’s a technical one

You cannot ship weekly if every launch is treated as a referendum on someone’s competence. The prerequisite for rapid iteration is institutionalising the right to fail, removing the stigma from a small, reversible, well-instrumented mistake.

Practically, in my experience, that means: launch narrow, measure honestly, and be genuinely willing to switch things off. The discipline is not in launching fast; it is in killing what does not work before it becomes a feature nobody owns and nobody uses.

If your team cannot name something they shipped and then removed in the last six months, you are not iterating. You are accumulating.

5. Predictive support: anticipate the calendar

Education runs on a predictable rhythm. Assessments, exam windows, enrolment deadlines and apprenticeship start dates are known months ahead. Very few providers use that.

Once we know what a learner is preparing for and roughly when, we can surface the right material before they go looking, and reduce the friction of the search entirely. The same logic applies to any provider: the user most likely to withdraw usually generates signals weeks before they disappear. Anticipating that, is not sophisticated AI. It is mostly a matter of joining up data you already hold.

Start with the friction you can already predict from the academic calendar before investing in anything more exotic.

The harder discipline is subtraction

There is one caveat running underneath all five points, and for me it is the most important thing happening in EdTech right now.

Generative AI has made it trivially easy to add, another feature, another integration, another automated touchpoint. It has also made passive consumption effortless. A learner can now get a plausible answer to almost anything without doing any of the cognitive work that actually produces learning. If our sector’s response is simply to hand learners more ways to consume answers faster, we will have built very efficient tools for not learning.

The defensible position, commercially and pedagogically, is on the other side: tools that make learners do something. Test themselves. Organise their own thinking. Practise.

In a market where every microscopic need has a thousand digital solutions, the real skill is no longer building. It is deciding what not to build, and having the nerve to remove what is already there.

That is a harder conversation to have in a strategy meeting than announcing a new AI feature. It is also, in my experience, where sustainable growth actually comes from.

By Paolo Muoio, COO at Docsity