Any beta group member predating May 1, 2026, paid individual subscribers, institutional users, OR the first 100 free users to utilize Assist.

We plan on launching the new Canopy Chrome Extension, replacing Annotate PRO, in August 2026. Beta users will have access to Assist until August 15th, 2026.

You can happily run both the production version of AP and the beta at the same time. 

Like many educators, we’ve tracked the advancements of AI with trepidation and wonder. Since the release of ChatGPT, we have experimented with AI and considered ways to provide AI-powered features in our solutions. We’ve listened to the conversation across global education, from early “AI detector” hopes to the embrace of AI in learning to a return to Blue Books. Bigger picture, we worry what today’s students will do for a living in 5 or 10 years, share concerns over AI’s environmental impact, and worry about the impact of AI on kids who will not know a world without LLM chat bots and agents.

11trees was founded and is run by educators. We’ve graded thousands of student papers and taught hundreds if not thousands of class hours in a variety of subjects. We are steeped in Dr. Grant Wiggins’ Authentic Assessment modality and believe deeply in the importance of writing and critical thinking skills. A world where students use AI to generate mediocre writing and submit it to educators who use AI to generate feedback is a dystopian future come true. We see no easy answers but we do have faith in human ingenuity to solve the energy consumption problem and believe rigorous education systems, anchored by critical thinking skills, is our best chance to evolve through the emotional, regulatory, and cultural challenges ahead.

We’ve been using AI ourselves, to code, to draft emails, to evaluate strategy, amongst many other tasks. And we’ve found some magical ways to put it to use.

Our breakthrough came when we realized that our “comment bank” solution, and particulary our investment in extending the idea of snippets of text to become Feedback Forms, provided a unique base for using AI. With Assist, your “comments” become instructions to AI that can be very prescriptive. The more context and detail you provide an LLM, the better its output. Just asking it to “grade a paper” will create a decent output although its style may not meet your needs or it may be too verbose. Develop a detailed prompt, that includes your rubric and the kinds of observations you use? And potentially example resources from your school to include where appropriate? With length and tone requirements? The LLM (we’re using Google Gemini) can draft a terrific response after reading the student’s work and even considering a Canvas rubric. Emphasis on draft. Canopy Assist drafts a response that you can review, edit, then add to a Google Doc (copy/paste) or Canvas (insert).

Things get even more interesting in Canvas (and soon, D2L). Canopy Assist can query a student’s previous Assignments and provide a summary of their progress. Assist can do the same, for a single Assignment, across an entire class, generating a learning outcomes progress assessment tailored for SACS or AACSB or your own internal structure. And we’re just getting started. We’ll be building on these ideas to add synthesis of exam data across a course to help educators gain insight into their students’ progress that would be unthinkable even 12 months ago.

And because institutional clients can create and share libraries of this content, efficiently, Canopy Templates (renamed from AP Comments) become the ideal vehicle to scale up efficient use of AI to serve education’s goals.

To do this work we need your help. How does Assist work when you apply it to a class with 200 students, each with Assignment feedback? Or a 10-page legal brief? We hope to hear about your experiences and descriptions of problems you need to solve that this technology might be able to support.

If you have the beta installed you’ll get automatic updates to the beta just like the production version of AP.

Absolutely. The beta version connects to the same data as the regular version of AP. Just log into Canopy with your regular credentials.

No. One of our goals is to understand the costs, for 11trees, in offering AI-powered features.

No. If you’re logged into Canvas or Google Docs, Canopy can support your work via secure connect to Canvas using Canvas’ APIs. Canopy does not store any of the information processed to provide you within insights into student engagement and feedback.

When using Assist’s AI features, our AI provider, Google Gemini, does not use the data to train its models.

We need to understand how well Canopy performs in real-world situations. For instance, a class with 120 students or similar. We anticipate expanding the options for querying Canvas as the beta continues. Your ideas deeply appreciated!

IgniteAI is very promising – a generalized chatbot to interact with Canvas. Canvas. Canvas also have “AI-Powered Grading Assistance” coming for SpeedGrader. We’ve used IgniteAI and it is impressive. Our analysis so far (and this is a fast-changing area of tech):

1. Using LLMs effectively, at least for the forseeable future, depends on writing detailed prompts – instructions. Annotate PRO pioneered the idea of “comment libraries” that can be complex, include form elements, and are easily shareable. Our shift to Canopy Templates reflects the importance of these instructions to getting great value from AI. Schools can license Canopy without Assist, if they choose to pay for IgniteAI, or utilize Canopy’s AI features.
2. IgniteAI is an additional cost for schools and pricing has not been shared publicly. Schools may opt not to license it and leverage Canopy Assist as a more targeted approach to leveraging AI. Or use Canopy without AI-powered Assist while utilizing Template sharing to simplify providing instructions to IgniteAI.
3. When it comes to drafting feedback on SpeedGrader submissions, Canopy Assist is already available (not coming soon, like Canvas’ feature) and will generate more personalized, specific feedback because of its use of Templates to instruct the AI. For instance, a school could share a library of such prompts, including detailed lists of school-specific resources that Assist can use to specify actions for the student. Canvas’ Auto-Evaluate feature scores a rubric (so assumes a rubric) and drafts criteria-level responses but doesn’t generate a narrative summary. Its only inputs are the assignment description and the rubric. Canopy Assist uses the assignment description (if available), rubric (if available), and the detailed instructions provided via the Canopy Template to draft a narrative response. Canopy Assist can also summarize previous feedback provided to the student to inform the current evaluation.
4. Canopy Assist is available outside of Canvas. The beta already supports Google Docs, D2L Brightspace support is coming soon, and Microsoft Word will be in place by August of 2026.

Institutional clients will have on/off control of Assist at the institutional level once we release the Production version.

While not yet available, we have architected Assist to support custom models and accounts at the institutional client level once Assist enters production.