Colleague AI

"Personalized learning" has appeared in approximately every EdTech pitch deck, conference keynote, and product homepage since 2015. It has been overused and without depth.

So let's be specific about what it means when Colleague AI says personalized learning: the ability to customize learning for any learner, at scale, every day, every week of the school year.

This means the lesson plans generated simultaneously across multiple Lexile levels — for students in a single class by their level, at once. It means IEP goal language drafted in the right terminology. Student tutoring adapted to each learner's level, in 50+ languages. Accommodation and modification language built from teacher notes. Not "personalized" as in one setting you toggle. Personalized as in: the platform does different things for different students, in real time, without asking the teacher to do it twice.

An 1816 dinner party. Mary Shelley. The birth of Frankenstein. And a group of Estonian students who lived it — role-playing the attendees, arguing in character, then stepping back to discuss what they'd just experienced.

Estonia has woven AI into classroom learning in ways that are genuinely educationally rich: not AI doing work for students, but AI creating conditions for deep engagement that couldn't happen otherwise. More here: https://www.wsj.com/tech/ai/estonia-schools-chatgpt-9ff76cc7

The Frankenstein dinner party conversation assignment is a pedagogical activity designed for the modern world, made possible by AI that can hold a character and adapt to student responses.

Colleague AI enables educators to generate interactives, assignments, and AI discussions such as a student discussion with Mary Shelley on Frankenstein. This is learning that meets students where they are, in formats that make them think harder, not less, avoiding cogntiive decline or outsourcing.

Vygotsky's Zone of Proximal Development — the space where a student can learn with the right support — has been foundational to learning science for decades. What it hasn't been is practical at classroom scale. Knowing a student's ZPD matters; having time to plan and build for 30 different ZPDs doesn't happen.

Colleague AI helps you identify where each student is, generates differentiated activity plans from that starting point, and then produces the actual interactive materials - a complete instructional sequence, grounded in learning science, ready to teach.

The conversation about education data science is accelerating — and the schools that understand it earliest will build the most durable AI strategies.

Last week, Dr. Lief (pictured, left) joined researchers and practitioners at the Stanford Education Data Science Conference, where the field is working out some genuinely hard questions: how do we measure learning in ways that actually inform instruction? How do data systems in schools serve teachers rather than just track them?

These aren't abstract questions for Colleague AI — they're the design foundation.

Recently, our CEO Dr. Min had the honor of joining a thoughtful and wide-ranging panel that brought together a diverse group of voices — tech company leaders, educators, investors, and academic experts — to explore how artificial intelligence is reshaping education today and in the years ahead.

Moderated by Olympia Trumbower, the conversation featured four panelists including representatives from Microsoft, a tech curriculum leader from Bellevue Schools, and an academic and nonprofit leader — all focused on navigating the opportunities and challenges AI presents for K–12 learning.

One of the things educators LOVE about Colleague AI's platform is that all of the interactive activities and assignments generated are accessible - friendly to learners who need accommodations.

Just click the accessibility man and a host of options appear!

No other platform can generate interactive assignments and activities in 5 minutes, let alone make them accessible.

The Center for Reinventing Public Education wrote an excellent article recently discussing what most K-12 AI tools or platforms miss:
https://crpe.org/getting-beyond-the-lightbulb-stage-why-ai-is-not-yet-transforming-education/

Some highlights:

Most AI tools are standalone and serve 1 function. Colleague AI is not.

Most AI tools have a "weak grounding in learning science" - this is why we spent years with education researchers and learning science professionals building Colleague AI - BECAUSE PEDAGOGY MATTERS!

Most tools are not aligned with education’s real challenges. Teacher productivity cannot be the goal in order for AI to have a positive impact. Colleague AI asks "what are your districts strategic priorities" and then we help you customize the platform accordingly and adjust training accordingly.

Not all AI tools are equal.

Getting Beyond the Lightbulb Stage: Why AI Is Not Yet Transforming Education – Center on Reinventing Public Education New AI tools are entering the ed tech market every day. But how are they actually playing out in schools and classrooms? Drawing on semi-structured interviews with more than 50 stakeholders across ed tech, philanthropy, policy, teaching, and advocacy, this brief identifies the gaps between what AI d...

Here's a real thing that happens when you train teachers on Claude: you hit the credit limit mid-session and get locked out for five hours. In the middle of a 4–6 hour professional development day. With a room full of educators who just got excited about AI.

Colleague AI was purpose-built for K-12 — not adapted, not retrofitted.

It already has agents designed for educators, no timed usage limits that can derail a training session, and it's priced for district budgets, not enterprise contracts.

When you're investing in teacher PD, the last thing you need is the platform tapping out before lunch.

Schools don't run on one job title. They run on the teacher who stays late, the counselor who notices, the custodian who has the building ready before sunrise, the bus driver who knows every kid's name, the front office that fields a hundred questions before 9am, and the leaders holding it all together.

To all of you — and everyone in between — thank you. Rest this long weekend. Recharge. You've earned it.

Our latest research is now live on arXiv: "Generative AI in K-12 Classrooms: A Midyear Implementation Report."

Produced jointly by Colleague AI and AmplifyLearn.AI at the University of Washington, the report looks at how teachers actually used AI across 12 school districts during the first half of the 2025–26 school year. These weren't cherry-picked pilots — the districts range from a few thousand students to thirty thousand, and span rural, suburban, and urban communities. The analysis combines platform usage data with district-provided administrative records to move past anecdote and into evidence.

This is the kind of work the field needs more of: transparent, mid-implementation data on what teacher AI adoption really looks like at scale. We're sharing it openly because districts deserve to make decisions grounded in evidence, not vendor claims.

Read the full report: https://arxiv.org/abs/2605.16277