5 Tools We Believe In from the AI Show 2025

What could possibly go wrong when an AI takes your drive-thru order?

Dozens of TikToks that surfaced showing frustrated customers; In the one above, the customer racked up almost $300 worth of chicken nuggets.  Another showed a customer pleading for the AI not to add bacon to his ice cream. By mid-2024, McDonald’s pulled the plug.

‍

The Wendy’s robot, unveiled a few months ago, represents AI implemented more thoughtfully.  It adds a human override if, say, someone orders an off-menu item. It’s a small tweak, but it’s the difference between going viral and passing the test. My guess is that McDonald’s was a little early, partnering with IBM back in 2021 before the AI technology was ready, while Wendy’s worked with a fully fledged model (not that recent models are better at the whole hallucinating thing).

‍

This whole situation is emblematic of where AI in education is today: the hype usually exceeds the delivery. 

• Try this AMAZING avatar tool… that looks creepy and unconvincing. 
  ‍
• AI will write ALL YOUR EMAILS… and miss the nuance every time.
  ‍
• Use OUR AI PLAGIARISM CHECKER… with a 49% success rate (and false positives to boot).

‍

So in education, where does AI hype meet reality? I found out at two conferences last month: the AI Show in San Diego and the AACOM conference in Arlington, Texas, where I got to play with prototypes pushing the edges of what AI can do.  The following lists the most impressive tools from those conferences, along with our (hopefully clear-eyed) assessment of what’s hype and what’s real.

‍

BoodleBox: One Workspace, Every AI Model

The hype: In a landscape where AI tools often operate in silos, BoodleBox emerges as a unifying platform, bringing together top-tier AI models into a single, collaborative workspace. If you want to teach with multiple AI tools — which is necessary at the moment, since their capabilities evolve quickly — Boodlebox lets your students use the right tool for the right task under one subscription.

Pros:

• Unified Access to AI Models: BoodleBox integrates leading AI models, including ChatGPT 4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, LLAMA 3.1, and Pi.
  ‍
• Collaborative Environment: Allows students to work together on the same AI interaction, something they can’t do with ChatGPT or Claude without physically working at the same computer.
  ‍
• Customizable AI Bots: Access to over 1,000 pre-built bots or users can build their own (with their own knowledge base).  OR you as their teacher could make a bot that could then be shared with the class, something that’s tricky with ChatGPT and Claude at the moment.
  ‍
• Pricing: Free version available;  Pro version is $20/user/month

Cons:

• Lacks latest features: In our tests, Boodlebox did not have access to the newest features of ChatGPT or Claude.
  ‍

• Rate limited at higher search volumes.  This isn’t new to power users of AI, as certain plans will limit the amount of deep research or interactions with higher models.  That’s true with Boodlebox too, and their limits sometimes are more restrictive than the AI tools themselves.
  
  

Does it live up to the hype?  We’re not giving out many of these yeses on this list, but this one has immediate application in the classroom and is the only tool on this list where we immediately purchased a subscription in the aftermath of the conference.  This tool solves a problem of sharing AI knowledge and results across our workspace.

Swivl M2: Your AI-Powered Co-Teacher

The hype: Imagine a teaching assistant that observes your lessons, offers real-time feedback, and steps in to support instruction — all without needing a coffee break. Meet Swivl M2, which looks a little like what you’d get if you attached an iPad to some PVC pipes.

Most AI tools at the show were student-facing, but this one’s for teachers.  Ever wonder how you could improve your teaching without your job being on the line? Swivl can do that, offering suggestions about where you could change your approach to align with the latest standards.  And because Swivl listens to you, it can hop in and continue teaching if needed. Now if it can only keep those unruly students from acting out…

Pros:

• Real-Time Feedback and continuous improvement: M2 provides immediate, actionable insights during lessons, helping teachers adjust pacing, engagement strategies, and instructional clarity after each lecture
  ‍
  
  
• Co-Teaching Capabilities: Need to step out or address a student’s needs? M2 can seamlessly continue the lesson, ensuring continuity and minimizing disruptions.
  ‍
  
  
• Non-Intrusive Implementation: Since M2 operates without direct student interaction, it’s easier to integrate into classrooms.
  
  

Cons:

• Pricey — the business makes money off its hardware, which starts at $2000 per unit.
  ‍
• Potentially flawed business model — Won’t ipads with AI be able to do this in, like, six months?  I could see a version of TeacherGPT (to be clear, a thing I probably just made up) that could fulfill a similar role.
  ‍
• I didn’t get to see it in action — There wasn’t much by way of a full demo available at the conference, but you can watch it in action here.
  ‍

Does it live up to the hype?  A tentative yes. Had the price been a little more affordable for your AI enthusiast, I just might have bought one to try at home.

‍

TrueMark: Rethinking Writing with AI

The hype: Don’t you wish you could see which parts of a student’s writing came from AI and which parts were their own?

That’s the question TrueMark is trying to answer. Their platform isn’t another AI writing tool for students; it’s an AI-powered feedback system designed to help educators track, evaluate, and guide student writing.  It feels like Google Docs, but with the ability to replay the writing process so you can see how long it took students to write a draft honestly, or, ahem, copy and paste.

For students, TrueMark lets them compare their draft against the assignment rubric in real time, helping them see gaps, improve alignment, and understand how feedback connects to the rubric’s goals. It also pre-generates feedback comments mapped to the rubric, streamlining the grading process.

Pros:

• Tracks writing process: shows how the assignment was composed. Did they paste in a block of text? Did they edit or delete large sections? TrueMark shows you
‍

• Student-facing rubric check: lets students self-assess against the teacher’s rubric
‍

• Embedded writing tutor: provides AI guidance while students draft
‍

• AI grading engine: pre-generates feedback tied to rubric criteria
‍

• Secure writing environment: students work in a controlled, Google Docs-like interface

‍

Cons:

• Requires switching writing tools from Microsoft or Google which seems like a hefty ask just for the AI powered abilities
‍

• Aimed at English class only, which makes implementation tricky.  This isn’t a “whole school solution,” so how many different tools is one school  —  or one professor  —  really going to use with their students?

‍

Does it live up to the hype?  Yes, for its intended purpose.  It’s the innovation of being able to replay a student’s work on an assignment, though, that’s so innovative.  Here’s hoping the education versions of other tools take note.

‍

EverTutor: The Promise of AI Tutors

The hype: One of the most intriguing, and unsettling, tools at the AI Show wasn’t a chatbot, a grading assistant, or a virtual patient. It was a tutor. A tutor that watches you while you work.

AI tutors aren’t exactly new.  Khanmigo was one of the original use cases for AI in education.  And if you use much software as a service, you’ll now notice a helpful AI chatbot to make the entire application easier (and, potentially, charge you more).

The potential is obvious: more personalized learning, less wasted time, more data to guide instruction. But so are the questions: How much “watching” is too much? Who owns the data? And how do we balance individualized learning with privacy?

Pros:

• Real-time observation: The AI monitors the student as they work, analyzing errors and strategies
‍

• Adaptive content: Suggests next problems and lessons based on live performance
‍

• Personalized pathways: Tailors the learning sequence to the student’s strengths and gaps

‍

Cons:

• Not yet public: Demo version finalized week of the show; only recently rolled out on website
‍

• One subject so far: Only works for GRE problems at the moment
‍

• Are we okay with AI watching us in real time? Is an AI watching you work a bridge too far for this generation of students?  (my vote for the Tiktok generation is no, but how comfortable are parents having it watch younger students?)

‍

Does it live up to the hype?  Not yet.  But count us as believers for this genre of tools.  The promise is too tantalizing to ignore.

‍

OpenNote: Notes That Do More

The hype: 

What if your notes were, like, really smart?  OpenNote is the note-taking tool for science and math that promises to easily embed charts and graphs, to make quiz questions based on your note, and to pop in the personalized tutor armed with all of the knowledge of your notes (and whatever AI training model it uses) to further illuminate the concepts.  

‍

Pros:

• Make and embed beautiful graphics, essential for understanding math and science concepts
  ‍
• Adds web search results right into your notes so that you can further explore pedagogical materials (though the quality of this may vary).
  ‍
• Free version available.  Upgrade is $15/month.

‍

Cons:

• Doesn’t beat perplexity.  A quick comparison of meiosis and mitosis, for example, in both Opennote and Perplexity.ai yielded similar results.  Not that those are precisely the same use cases.
  ‍

• Early, early, early stage app.  The founders are dropping out of college to work on the app, so its performance isn’t quite what you’d expect from a more full-fledged app.
  ‍
• Imperfect results.  Simple queries to explain a concept or make a video returned with results that were not super useful.

‍

Does it live up to the hype?  No. I really liked the founders and think they’re on to something.  Specifically, smart notes seems like an ideal use case for AI, not to replace the essential act of taking notes in class but to use the notes you already take and supplement them with a variety of tools personalized to you.  

‍

Clinical Reasoning Tool: InSitu by ReelDx

We weren’t at the AI Show. But we were at the AACOM conference in Arlington, presenting our latest work with InSitu, a simulation platform that blends AI with something AI can’t fake: real patient video cases.

Where many tools rely on synthetic data or fabricated scenarios, InSitu’s power comes from authentic patient stories — captured with consent, anonymized with care, and delivered in ways that challenge students to think like clinicians.

‍

The hype: Work through standardized patients without having to hire actors! With an InSitu case, you get the realism of real patient video with the interactivity of a chatbot that sounds like the patient from the case.  Then the AI preceptor grades you on your interview and your empathy.

‍

Pros:

• 40 patients with different conditions and in different clinical locations
  ‍

• Students all work from the same patient and condition, helping to standardize their exposure to specific cases
  ‍

• Chatbot interface adds critical thinking.  Students don’t just choose the next text from a dropdown but have to come up with what to say and how to say it.

‍

Cons:

• No ability to perform a physical assessment on a virtual patient.
  ‍

• Grading remains formative, not summative.  But teachers can see what each student said in their interview at the very least.
  ‍

• Pricing starts at $69 per student per year.  No free option as of yet.

‍

Does it live up to the hype?  It certainly doesn’t replace standardized patients or the gold standard of clinical rotations.  But hopefully, that’s not the intended use.  Instead, this is a tool to simulate the realism of a clinic outside of the clinic.  As the AI behind it evolves, the realism increases.  Zoom-call style audio encounters are starting to crop up on tools such as Duolingo, and the same could be an ideal use case here, where you “meet” the patient from the video and then interview that patient, with an AI voice that sounds just like the patient from the video.

‍

Conclusion:

Keep an eye on these tools — and the broader category they represent. They’re the most market-ready, the most likely to change practice, and the ones that could stick… as long as they avoid going viral for the wrong reasons.

‍