Ativion Uses Amazon Nova Foundation Models to Power Its Generative AI Classroom-Management Solution

ContentKeeper is Ativion’s proprietary content-filtering service, which protects more than 10 million students in 80 countries. This proven solution remains at the core of the company’s offerings. When envisioning the next generation of classroom-support tools, the Ativion team saw an opportunity to build on ContentKeeper’s strong foundation. Enter StudentKeeper—a solution that is powered by ContentKeeper and that harnesses the power of generative AI to filter content, analyze access patterns, and deliver near real-time insights to teachers and administrators. “ContentKeeper is an exceptional filtering solution,” says Lionel Farr, chief technology officer at Ativion. “With StudentKeeper, we’re taking it a step further—turning that data into actionable insights that empower educators and elevate the learning environment.”

The data previously collected through ContentKeeper was substantial—and a potential gold mine of insights. Schools using ContentKeeper were creating 200–400 million data events per day. This data captured the ways that students were accessing content, the duration of the content access, and other student actions. Managing and filtering this extensive data could help engage students with learning and provide new capabilities for teachers. It could also uncover valuable insights for educators, identifying possible blind spots and improving the educational experience. “Beyond filtering, what schools needed was the ability to manage classrooms and get a real picture of the digital body language of every student,” says Farr. “And we needed to make teachers’ lives easier, because technology has made their work so much more complicated.”

Ativion’s team includes dozens of members with extensive teaching experience who know these challenges firsthand. “Managing a classroom of students using laptops is a real challenge,” says Travis Merritt, product operations specialist at Ativion and former teacher with 23 years of classroom-teaching experience. “Keeping students from getting distracted and going off task is like playing Whac-A-Mole. You can call inattentive students out, but they just go right back at it when you’re not standing in front of them.”

But Ativion’s goal wasn’t to increase classroom monitoring. “Blocking, flagging, catching kids doing things they shouldn’t—that’s not what we wanted to do,” says Channing Anderson, product manager for StudentKeeper at Ativion, who also has years of teaching experience. “We wanted to use the data that we had available to see what’s happening in the classroom and improve the experience for students and teachers. Yes, we wanted to watch out for students who are hurting themselves or impacting the learning of others. But on a higher level, we wanted to create a positive environment that rewards students for being focused and engaged and for using technology responsibly—and that makes it much easier for teachers to manage classrooms and nurture students.”

Cost was a key consideration as Ativion started building out StudentKeeper. The team had been experimenting and running pilot generative AI–powered applications for several months and was acutely aware of how costs scale as usage ramps up. They did not want additional fees to be an obstacle to adoption as they rolled out the new solution, so balancing the quality and cost of inference became a priority for StudentKeeper.

As Ativion was building its Academic Resource and Intelligent Assistant (ARIA), the AI-powered core of StudentKeeper, Farr and his team began experimenting with the versions of models offered through Amazon Nova Foundation Models that were announced in late 2024. “As we started exploring models, the light bulbs started going on among the team, who realized we could do a lot more with generative AI in StudentKeeper thanks to the quality, cost-effectiveness, and speed of using Amazon Nova Foundation Models,” says Farr.

During evaluation, the team found that the Amazon Nova Pro and Amazon Nova Lite models provided the multiple price-performance tiers and capabilities required for different use cases within ARIA’s multilayered AI functionality. “As an AI assistant, ARIA is all about answering requests for information,” says Farr. “I would specify system prompts, like ‘Tell me about the Great Wall of China,’ and Amazon Nova Pro was very good at providing structured feedback in the conversational formats that we care about for both students and teachers, who have very different needs. We found that Amazon Nova Lite was the right model for the more operational and tactical aspects of ARIA, such as quickly providing an ongoing ranking of student attentiveness. Its responses are very fast, helping teachers to make near real-time adjustments in the classroom. And we found both Amazon Nova Pro and Amazon Nova Lite were lower in cost than other models we evaluated for the same tasks.” In fact, Ativion found that Amazon Nova Pro delivered results 30 percent faster and Amazon Nova Lite was over 700 percent more cost-effective than comparable models.

Based on its evaluation and exploration of Amazon Nova Foundation Models, Ativion chose Amazon Nova Pro and Amazon Nova Lite to power a range of key generative AI use cases. “Using Amazon Nova Foundation Models provided the combination of accuracy, performance, scalability, and affordability that we needed for ARIA,” says Farr. “That’s why Amazon Nova Foundation Models became the foundation of StudentKeeper.”

The team chose Amazon Nova Pro to power the more sophisticated AI capabilities of ARIA and used Amazon Nova Lite as an efficient, lower-cost utility model. “Depending on what we’re doing and the use case we’re supporting, the solution will hop between models, mostly for speed and cost-efficiency,” says Farr. Ativion built StudentKeeper on Amazon Bedrock, a fully managed service that offers a choice of high-performing foundation models, which helps the solution to switch seamlessly between Amazon Nova Pro and Amazon Nova Lite, depending on the use case. The built-in security of Amazon Bedrock keeps sensitive student data secure, while Amazon Bedrock Guardrails provides safeguards that Ativion can configure for generative AI applications based on the use cases and responsible AI policies.

Amazon Nova Foundation Models power three core capabilities of ARIA: a safeguarded AI assistant, student insights, and teacher support. Amazon Nova Pro powers ARIA’s AI assistant, which helps students navigate AI safely while providing teachers with extensive support tools. When students use ARIA to explore topics and deepen their learning, Amazon Bedrock Guardrails provides safeguards to prevent inappropriate or unauthorized AI responses.

Based on each student’s searches and activities during class, ARIA uses Amazon Nova Pro and Amazon Nova Lite to gather data and calculate near real-time monitoring and attentiveness scores that show each student’s level of engagement with learning. These timely insights help teachers and parents know more about a student’s progress so that they can refocus a student’s attention when necessary. The models also flag potential safety concerns that need to be addressed proactively

ARIA uses the intelligent document-processing capabilities of Amazon Nova Pro to accelerate time-consuming tasks, such as preparing ongoing notes on progress and issues for dozens of students. Amazon Nova Pro extracts essential information, edits it for clarity and completeness, and automatically populates all required fields. “Using Amazon Nova Foundation Models, we make it easier for teachers to do their administrative tasks, so that they have more time to focus on delivering customized learning experiences, not doing busy work,” says Farr. In the past, teachers could create a handful of notes every day, often writing them after school hours. Using ARIA and Amazon Nova Pro, teachers can create dozens of accurate, complete notes in minutes, helping to quickly identify patterns and guide timely interventions.

Early users of StudentKeeper—which is in beta testing and due to launch in the summer of 2025—have responded positively to the capabilities and experience that it delivers. The EdTech industry has already recognized the innovative technology of StudentKeeper, which won a Tech & Learning Award of Excellence: Best of 2024 in the Secondary Education category.

Part of StudentKeeper’s appeal is that it makes sophisticated generative AI capabilities simple to use for students and teachers, with the responsiveness and reliability needed to deliver a compelling experience. “Amazon Nova Foundation Models are working behind the scenes to provide fast, reliable information, insights, and assistance,” says Farr. “But they are invisible to students and teachers, who just use ARIA as part of their classroom experience, because, at the end of the day, our goal is to use generative AI to empower teachers to be more efficient while helping students to learn better—all by keeping ARIA very simple for everyone to use.”

The broad input from teachers that helped shape the design and functionality of StudentKeeper was obvious to early users. “They can see that StudentKeeper was built for educators, by educators,” says Anderson. “We have people with education backgrounds throughout Ativion, from solutions engineering to marketing. And we poured all that energy and experience into creating the solution that we knew we needed to manage the classroom efficiently, to put the focus back on teaching, and to keep students engaged and safe.”

By building StudentKeeper using Amazon Nova Foundation Models on Amazon Bedrock, the Ativion team was able to deliver the quality and speed needed for ARIA’s diverse use cases. As Ativion prepares for StudentKeeper’s production rollout, Amazon Nova Foundation Models will help it to scale StudentKeeper globally in a reliable, cost-effective manner.

Looking ahead, the Ativion team sees StudentKeeper’s generative AI capabilities evolving and increasing. “We’ll continue expanding how we build on Amazon Nova Foundation Models and Amazon Bedrock to facilitate the teacher journey and enhance the student educational experience,” says Farr. “I truly believe that generative AI will be part of their lives forever, so bringing it to the classroom in the right way is critical to K–12 education and beyond.”