What teachers need from AI

Tools that are ridiculously simple, immediately useful, and pedagogically sound.

I took time off recently and spent some of it working on a personal project of mine: an AI-powered lesson-building tool called Insta~Lesson.

To be honest, a general-use AI tool is probably the last thing I should be spending my time on. For one, I have young kids and the weather is getting nice. For another, I’m deeply engaged in my work at Modern Classrooms Project, which includes using AI to help teachers implement high-quality curriculum more effectively. Finally, the market for teacher tech tools already feels saturated, especially with tools that use AI.

Yet despite this saturation, and despite two-plus years of excitement about the ways in which AI can transform education, I find that education hasn’t changed much at all. As surveys have shown, most teachers aren’t really using AI to begin with. And despite rapid technological advances and widening achievement gaps, an average classroom today doesn’t operate all that differently from an average classroom ten or even one hundred years ago.

In some ways this makes sense. Education is too important to overhaul every time a shiny new technology hits the market. Teaching is, has always been, and always will be a fundamentally human endeavor.

Still, there is room - and opportunity, and need - for improvement. Our students’ rates of proficiency are abysmal. The traditional, one-lesson-per day model of instruction fails to meet millions of learners’ needs. Teachers are burning out fast.

I believe that tech in general, and AI in particular, can help address these challenges. So why haven’t AI tools made a real impact?

I think the answer is simple: most AI tools for teachers aren’t yet meeting teachers’ daily needs. These tools too hard to use; their outputs aren’t sufficiently useful; they fail to support effective pedagogy. It’s a new market, yes. But the tools in it don’t yet provide teachers or students the support they require.

Of course, it’s easy to criticize. And I’m a builder at heart, so I never like to criticize without demonstrating - at the very least to myself - that a better way is possible.

Which brings me back to Insta~Lesson. I’m tired of waiting for the right tool to be built, so I am trying to build it myself. And in this post I want to suggest:

1. Where current AI tools for teachers fall short.

2. How Insta~Lesson attempts to address those shortcomings.

And I’ll be the first to admit that Insta~Lesson isn’t a perfect product yet. In many ways, and despite Teaghan O’Brian’s incredible work engineering it, Insta~Lesson is really just a proof of concept! But I think it proves something worthwhile: that a purpose-built AI tool really can make teachers’ and students’ lives much easier.

Let’s take a look.

What I talk about when I talk about teacher-facing AI

What I’ll discuss here is the limited subset of AI tools that are designed to help teachers create learning materials for students. As someone who works in teacher professional development, those are the tools with which I’m most familiar, and also the AI tools I tend to believe have the most immediate potential.

I believe these tools are so promising because, as a former teacher myself, I know that:

1. Teachers spend most of their time delivering or facilitating learning experiences.

2. Planning effective learning experiences takes a lot of time.

And while I’m skeptical of many educational use cases for AI - student-facing chatbots, for instance, or automated feedback generators¹ - I’m confident that AI can help teachers generate effective lesson materials. Insta~Lesson already does that! I’m just not convinced that any of the many other tools on the market do (yet).

Here’s what I think teachers actually need.

Need 1: Simplicity

Teachers (like everyone else) prefer tools that are simple to use. The daily work of teaching is already hard enough! The last thing an exhausted teacher needs is another complex platform or system to master.

Fortunately, AI promises ease of use. Enter a prompt, let an LLM do the work, and get a mostly finished product in seconds!

Except… it’s not that simple.

Take ChatGPT. It’s interface seems extremely user-friendly:

ChatGPT: One tool to do anything. Literally.

Using ChatGPT effectively, however, is an art. You need to know how to write a good prompt, how to provide feedback ChatGPT can interpret, and how to minimize hallucinations. That’s why there are courses offered and books written about effective prompting, not to mention companies hiring full-time prompt engineers.² How many working teachers have the time to learn all that?

On the other end of the spectrum are platforms designed explicitly for teachers. These tools help teachers create all sorts of classroom-specific resources, such as:

• Worksheets

• Assessments

• Rubrics

• Slideshows

• Lesson plans

• And so on.

Some of these tools have literally hundreds of templates that teachers can create!

So while I appreciate that these tools are tailored to teachers’ use cases, I also find them overwhelming. I’m never sure what resource to create, or how to create any particular resource most effectively. And while I’ve heard great things about the PD that many AI companies provide teachers, I think great products should be easy to use without the need for any PD at all.

That’s why Insta~Lesson asks teachers to begin with three simple, well-defined inputs: the topic of an upcoming lesson, their students’ grade level, and an optional learning standard.

The user interface at insta-lesson.com

There’s a lot happening under the hood here: when a teacher enters these parameters, that information is run through carefully vetted prompts which are likely to generate high-quality outputs. (We’ve taken the courses on prompting so teachers don’t have to.) But getting started - and then moving through a step-by-step lesson-building process - couldn’t be simpler. You enter a topic and you’re off to the races from there.

Need 2: Usefulness

Teachers’ time is extremely limited. When teachers invest that time into using an AI tool, they need to get something extremely useful out.

So say you invest the time to learn a tool like ChatGPT, and figure out how to coax decent materials out it. Then what?

One of the biggest challenges for teacher-facing AI is what Dan Meyer calls its last-mile delivery problem. Even once you’ve prompted AI to give you what you want, you still need to make sure it will work for the students you know best - and then get that efficiently into your students’ hands.

Unfortunately, most AI tools don’t make that easy. Most of the time, they simply output plain text, which you must then copy and paste this into a format you can use with students.³

And a full lesson typically requires several different resources: a warm up, a slide deck, practice work, an exit ticket, etc. So even once you’ve formatted any one AI-created resource, you still need to generate and reformat several other pieces of the lesson. That’s a lot of work! So much, in fact, that it might be easier not to use AI at all.

(And please don’t suggest that I can use a lesson-plan generator for this purpose! As any teacher knows, a lesson plan is a far cry from a classroom-ready lesson.)

Insta~Lesson, on the other hand, builds fully formatted - and fully editable - Google Docs, which teachers can give directly to their students, either digitally or on paper.

A ready-to-share, two-page lesson created by Insta~Lesson - with plenty of blank space for student work.

And that’s not all Insta~Lesson creates. For any lesson topic, Insta~Lesson will create (all at the same time):

• Warm Up (example)

• Student-facing worksheet (example)

• Extension activity (example)

• Exit Ticket (example)

• Teacher guidance (example)

• Lesson slides (example)

Now THAT is a complete and ready-to-share lesson. And it’s all fully editable too.

Need 3: Pedagogy

The needs above are practical. But what teachers need more than anything, in my opinion, is a philosophy of instruction that really works.

And here is the biggest problem with most teacher-facing AI tools. These tools rarely support effective pedagogy.

After all, AI “learns” from what already exists. Large language models are trained on data from the Internet, which reflects the state of the world as it currently is.

So it’s clear that AI can competently replicate the kinds of resources that are widely used in traditional classroom settings. But it’s also clear from decades of NAEP scores that, for the majority of American students, the traditional model doesn’t work. AI can surely facilitate ineffective instruction, but whom does that really serve?

Insta~Lesson, on the other hand, is built around evidence-based pedagogy. Decades of research have shown that students learn at different paces, and that mastery-based learning helps all students succeed. (Good summaries here.) So Insta~Lessons are explicitly designed to foster learning experiences in which students:

• Access direct instruction through high-quality instructional videos, freeing teachers up to provide targeted support in class.

• Develop understanding at their own paces, usually in collaboration with peers.

• Have the chance to prove their understanding before they advance.

The directions for every Insta~Lesson. Students learn; teachers support.

This is AI in service of good teaching: the tool makes the pedagogy possible. That kind of teaching - and therefore that kind of tool - is what our young people need.

Miscellaneous Teacher Needs

I’m almost finished, I promise. And I’ve covered what I see as teachers’ major AI needs above. But if I’ve still got your attention, I want to flag two other things teachers need from AI - and how Insta~Lesson addresses them.

1) Ease of getting started.

Teachers hate creating unnecessary accounts. Most already have to deal with a patchwork of logins - SIS, LMS, etc. - and the focused time they have at a computer on any given day is limited.⁴ So having to create an account (and often complete some kind of user onboarding) before actually seeing what an AI tool can produce is infuriating. As often as not, teachers won’t bother.

Insta~Lesson, on the other hand, asks teachers for the bare minimum - their email addresses - and only to send teachers the Google Documents it generates. (While student-facing worksheets are generated in seconds, the teacher guidance and lesson slides can take up to two minutes.) If teachers want to sign up for accounts later, they can. But the goal is to get teachers the materials they need as quickly as possible.

2) Unlimited free resources.

If you’re serious about empowering teachers, you won’t impose limits on how much they can create for free. But many AI tools do, which strikes me as counterproductive. If you want teachers to depend on your tool, let them use it as often as they’d like - for free! And don’t make them feel like they’re missing out by using the free version.⁵

Can tools have premium features for a cost? Of course! Business is business, and Insta~Lesson has paid features too. But the tools which will make the biggest impact are those which encourage unlimited free use. And I’m proud to count Insta~Lesson as one of those.

Where Things Go from Here

Is Insta~Lesson a perfect product? Not yet. It has barely been released! But as I hope I’ve shown, it already improves on current AI tools in meaningful ways. And we’ve only been working on it for a few short months, I can only imagine what’s ahead.

A lot of things feel possible as I write this article. It’s possible that Insta~Lesson will take off. It’s possible that another AI company will listen to what I’ve written, incorporate my ideas, and render Insta~Lesson obsolete. It’s that Insta~Lesson isn’t actually all that useful in the first place.

But if any AI tool can save teachers time in building lessons - whether that tool is Insta~Lesson, something new, or a better version of something that already exists - it will be good for the world. And even if no one takes any of what I’ve written here into account, at the very least it helps me justify Insta~Lesson.

In the meantime, though, I’m going to go enjoy the nice weather.

1. I’m not particularly familiar with most student-facing tools, and I’ve been impressed by Snorkl, EdLight, and Coursemojo. While my sense is that students would generally rather learn from one another and their teachers than from AI bots, I recognize the value in outsourcing some student-facing work to technology, and realize I may be wrong here.

2. If you’re looking to learn more about prompting, I thought this book by Catlin Tucker and Katie Novak was excellent. But I know a lot of teachers just won’t have the time.

3. Some AI platforms for teachers offer export options, but this is often a premium feature.

4. This is part of the reason why Clever is so brilliant.

5. I hate trying new tools and seeing the features I really want to use grayed out, with some cheery upsell to the paid version. Teachers already deal with so much - do we really need to remind them that they’re getting subpar products too? Insta~Lesson has additional features for Partner Educators, but you won’t see anything grayed out on the free version.