Redefining instructional support with generative AI: How Instructure built IgniteAI on AWS

Increasing numbers of teachers report experiencing stress and burnout in their jobs. Many are even considering leaving the profession, especially those with more experience in the classroom. One of the leading causes, according to a 2024 RAND Corporation study, is the amount of administrative work that pulls educators away from teaching.

To address this challenge, Instructure, a leading education technology (EdTech) company, has announced the release of IgniteAI, a new suite of generative artificial intelligence (AI) solutions designed to simplify classroom administration. Developed in close collaboration with Amazon Web Services (AWS) and built on Amazon Bedrock, IgniteAI and its agentic AI feature, IgniteAgent, are deeply integrated into Canvas, Instructure’s learning management system (LMS) used by over 2.4 million teachers worldwide.

This blog post explores how the collaboration between AWS and Instructure is redefining instructional support and taking education technology to the next level.

Reducing time-consuming tasks for educators

Today’s educators are constantly under time pressure, especially in the wake of the COVID-19 pandemic, which left many students needing additional academic support. Some of the most time-intensive tasks include managing Individual Education Programs (IEPs), adapting lesson plans to meet diverse learning needs, and grading student work.

While LMS platforms like Canvas aim to streamline routine tasks, they can sometimes feel like just one more thing to learn or complete in an already overwhelming workday. A peer-reviewed 2024 study by Auburn University found that many schools implement LMS solutions without retiring old procedures, meaning the LMS becomes an added task rather than a replacement for manual processes.

“The last thing we want is for instructors to see technology, or a platform like Canvas, as a burden,” said Zach Pendleton, chief architect at Instructure. “So we began looking at where instructors were spending time in Canvas and asked: What can we simplify or automate so they can spend more time with students?”

An intelligent assistant built to support teachers

To reduce that burden, Instructure’s team began exploring how generative AI and agentic workflows could make Canvas more intuitive. By applying natural language processing, they could enable educators to ask the LMS what they needed in their own words, allowing the system to handle the rest.

For example, when a student needed extra time on an assignment, teachers previously had to navigate multiple screens. With IgniteAgent, teachers simply type a prompt, such as “Sally needs an extension,” and IgniteAgent updates the due date automatically. Tasks that once took multiple steps to complete can now be done with a single conversational prompt.

Other streamlined workflows include:

• Creating rubrics for assignments.
• Summarizing course discussions.
• Generating assignments or quizzes from uploaded content.
• Copying course materials.
• Drafting announcements in institutional tone and pedagogy.

IgniteAgent achieves this by seamlessly integrating into the more than 500 APIs available through Canvas’s open ecosystem. Importantly, teachers can preview the steps the agent plans to take and verify them before execution. This keeps a human in the loop and prevents the agent from taking unwanted actions.

“It’s important for our customers to understand what steps the agent is going to take, and which APIs it will call,” said Narine Hall, principal AI product manager at Instructure. “Educators really appreciate that transparency. It ensures the agent doesn’t take any random actions.”

Built responsibly with transparency and control

Fostering responsible use of AI is a top priority at Instructure and was the driving force behind its emphasis on transparent insights in all of its products. IgniteAgent extends this commitment to responsible AI by taking advantage of Amazon Bedrock’s regional hosting capabilities, allowing sensitive student data to remain in-region and be protected from being used for model training or development.

“The agent never acts on its own,” added Pendleton. “It always takes direction from teachers, making it more of an instructional assistant than an autonomous voice in the classroom.”

Scaling quickly with support from AWS

Instructure’s collaboration with AWS began 15 years ago, when the company made the decision to host Canvas on the AWS Cloud. So when it came time to scale IgniteAgent from concept to beta, AWS was the natural choice. In early 2025, Instructure engineers connected with the AWS Product Acceleration team. Within six months, they had a working beta ready for classroom testing.

The AWS team advised on effective use of large language models (LLMs) and helped refine prompts. “This is cutting-edge technology, and it was helpful to have access to subject matter experts at AWS who are deeply familiar with this space,” said Pendleton. “They were also able to deliver pieces of functionality that we didn’t have the expertise for in-house. That hands-on-keyboard time was very valuable.” IgniteAgent is powered by Anthropic’s Claude via Amazon Bedrock, Anthropic’s frontier, state-of-the-art large language models that offers important features for enterprises like advanced reasoning, vision analysis, code generation, and multilingual processing.

The team credited both the collaboration and Amazon Bedrock’s capabilities with accelerating development. “Building on Bedrock was incredible. It really accelerated what we could do in the timeline that we had,” said Hall.

Designed to support student-centered teaching

During the beta release, instructors in both K12 and higher education institutions tested IgniteAgent’s ability to solve real-world problems.

One teacher asked IgniteAgent to build a quiz from a PDF teaching guide. The system generated multiple-choice questions, aligned to instructional objectives, and then invited the teacher to edit. Another uploaded a research paper and received a structured critical-thinking assignment, complete with grading criteria and an optional rubric.

Teachers described the experience of using IgniteAgent as empowering and appreciated that it maintains a posture of providing support, not serving as a substitute. “It’s like having an instructional assistant who never tires,” said a beta tester.

Powering the future of personalized education

IgniteAgent is the result of cross-functional collaboration across instructional designers, engineers, researchers, product managers, and more. Their goal? To reduce the amount of time educators spend on administrative tasks so that they can focus on their students. The origins of IgniteAI lie not in showcasing AI, but in addressing the real, daily challenges that teachers face. As demand grows for personalized learning, generative AI tools like IgniteAgent will help educators support struggling students and offer more to advanced learners, without increasing their workload.

By freeing teachers to focus on instruction, IgniteAI represents a step forward in building a more efficient, responsive, and student-centered educational system.

To learn more about how AWS helps EdTechs build, deploy, and scale solutions that facilitate student success, contact us today.

Read related stories on the AWS Public Sector Blog:

• Unlocking student success with generative AI: How Panorama Education built Solara on AWS
• 6 EdTechAI trends: How artificial intelligence is reshaping education
• From vision to reality: Inspiring your educational institution’s first steps in AI with AWS
• Anthology launches omnichannel student services transformation with Amazon Connect