What AI tools could we have had in 2025 but for some reason didn’t get? We’ll look at five obvious candidates that somehow missed the boat.

Note that when we say “from the last year,” we mean “what could have been done last year but wasn’t.” We’re not talking about big multi-year projects like robot teachers or personal tutors. We’re talking about five things that could reasonably have been developed, marketed and become popular some time in 2025.

Also, it’s hard to prove a negative, so we’re not saying these things never happened. It’s that they didn’t break into the mainstream.

Finding the Right Thing Is Still the Hardest Part

It’s still hard to find the right fit — podcast, video, article — that helps us learn something we want to learn. AI is always trying to explain or describe something when it should be guiding us to the really good resources that are already out there.

Why does this matter? There’s a difference between just describing or explaining, which is what AI does, and actually teaching, which is what a well-designed resource does. AI has learned how to tell, but not to show.

Why didn’t it happen? The business model doesn’t really work. Finding the best resource doesn’t really showcase AI’s capabilities, and it doesn’t keep you on the platform paying for tokens.

What we lose when we don’t have the right fit is an immediate and direct benefit from AI that helps us learn when and how we need to learn. We also lose the feeling that we human creators can work together with AI instead of in competition.

We Still Don’t Really Know What We Have

An AI can look at a large database and find all the green things, or all the German things, just by looking at the data. We can’t. We need labels. AI would be great at labelling and organizing all the files on our computer or in our organization.

Why does this matter? Labels and other metadata make it easy for us to comprehend large bodies of information, and to use it to do things like prepare lessons, articulate a concept or find just the right thing for francophone eight-year-olds in Northern Ontario.

Why didn’t it happen? Auto-indexing probably did happen inside some large enterprise systems, but it’s the sort of thing that’s hard to sell as a standalone product. And understanding just what to label is a difficult problem to solve for individuals and small enterprises.

What we lose when we don’t have auto-indexing is a way to better understand what our own digital world looks like as a whole. Who do we talk to most, what do we talk about most, what’s our personal go-to for problem solving? Automated labelling lets us see the answers to all of these.

Summaries are Easy; Understanding is Not

It has happened to all of us: a new government report comes out, or a scientific study or a compelling new book. We could take the time to read it, but it’s going to take a lot of background knowledge to really understand what it means.

There are more and more different fields of expertise in the world, requiring deeper and deeper background, and it’s impossible to keep up with everything. What we need isn’t just a summary, but something that explains it to us in a way that makes sense.

Why didn’t this shift to understanding happen? Explaining requires an understanding of what we know and tailoring the description to that. But how does an AI learn what we know? We have to interact with it a lot, and the AI has to remember things about us. Most of us aren’t ready for that yet.

Without understanding, what we lose is a way to get really useful summaries from an AI — and not just generic summaries that simplify things in the blandest ways possible so that we think we should be following along but we really aren’t.

Online Learning Still Struggles to Create Real Groups

It’s a great practice to break the class down into groups for hands-on activities, but it’s not so easy to do when the class is online, very large or both. Automated group formation seems like it should have been low-hanging fruit for AI in 2025, but it never really happened.

Why does this matter? Ideally it would be helpful to have a good balance of skills, interests and attitudes, but it would also be sufficient just to create groups that use the same technology or can meet at the same time. It would be enough even to form groups only from active participants in the course, saving students from being the only one in a group of five to attend the discussion.

Why didn’t it happen? It’s hard to get this right, although with the right data it’s trivial for AI. That’s where the difficulty lies. In a world where people aren’t willing to fill out a Doodle meeting planner, they’re even less likely to let an AI share when they can be online and what sort of skills they’ll bring to the table. And schools aren’t willing to snoop in students’ personal calendars or LinkedIn profiles.

What we lose without real groups is a way to move large-cohort and online learning from one-way content delivery sessions to genuine opportunities for collaborative learning, with all the social and educational benefits that entails.

Students Are Left to Integrate Everything Themselves

One of the greatest challenges for a student balancing several classes with work and home obligations is to keep track of what they’re supposed to do and when they’re supposed to do it. It should be trivial for an AI to collect the information from a student’s calendar and course outlines and offer a single integrated dashboard. But we didn’t see that in 2025.

Why does this matter? It probably won’t stop students from waiting until the last minute and pulling an all-nighter, but it can prevent obligations from getting out of hand during busy periods. Just being able to stay on top of what needs to be done reduces stress and helps a student feel more in control.

Why didn’t the integration happen? Many apps already exist that offer students task and calendar management, but these typically depend on students inputting data manually. It’s hard to make the leap to automated data collection, partly because of privacy and security concerns and partly because of the reluctance of providers such as learning experience platforms to make data available to other applications.

What we lose when we don’t have this integration is a chance to give students a single place where they can get organized along with a relatively painless way of achieving it. Without this opportunity, they don’t get to experience the real benefits of effective time and task management.

Maybe next year…