AWS Public Sector Blog
Boise State University builds campus-wide AI platform on AWS and cuts per-user costs by 80%
Boise State University serves 30,000 students and 1,400 faculty, making it Idaho’s largest university. By 2024, students and faculty across campus were using artificial intelligence (AI) tools, but adoption was fragmented—some paid for subscriptions to off-the-shelf solutions, while others used no cost versions. For a university pursuing R1 research status, this patchwork approach wasn’t sustainable.
The university faced three barriers to scaling access. Individual licenses would cost over $7 million annually. Family Educational Rights and Privacy Act (FERPA) regulations prevented using commercial tools that might train AI models on student data. And vendor lock-in would limit flexibility as technology evolves. Working with Amazon Web Services (AWS), a three-person Boise State team built boisestate.ai, a custom AI platform, in just six months. The solution gives the entire campus access to multiple models while cutting per-user costs by more than 80%.
Balancing cost, compliance, and equity
Commercial AI subscriptions posed interconnected challenges for Boise State. As a state university operating under FERPA regulations, Boise State couldn’t use commercial tools that might train AI models with student interactions or institutional data—a compliance barrier that eliminated many viable options. Meanwhile, students who could afford individual subscriptions gained access to more powerful AI capabilities than their peers, creating a two-tier system that disadvantaged students based on financial means.
Vendor lock-in posed yet another concern. Committing to a single vendor’s approach would constrain how faculty and researchers could apply AI as the technology evolves. “We want to maintain our freedom to choose and swap best-fit models,” said Phil Merrell, executive director of cloud architecture and innovation at Boise State.
The economics alone made commercial subscriptions prohibitive. At $20 per user per month, providing premium AI service subscriptions to the entire campus population would total $7.2 million annually. “That’s just unsustainable at scale for us,” Merrell said. While lower-cost options existed, these typically required significant compromises—access only to older, less capable language models or limited functionality that would undermine their commitment to providing cutting-edge resources to all students.
Merrell knew the university needed a different approach. His team began exploring how to build a solution on AWS that could deliver affordability and advanced AI access without sacrificing compliance, equity, or flexibility.
Working with AWS on a consumption-based model
Boise State had used AWS for eight years when it started exploring AI solutions in December 2024. This existing relationship proved valuable—AWS doesn’t require large upfront capital expenditures. Instead, AWS operates on a consumption-based model where customers pay only for what they use. For a university watching every dollar, this meant the team could start small and scale up without committing to fixed subscription costs. “That model works really well with higher ed, where we don’t have to make these huge upfront capital expenditures,” said Merrell.
Working closely with the university, Kat Ciovacco, an AWS account manager, helped Boise State map technology solutions to desired outcomes. “We’re always focused not on what we want, but on what the community needs in order for us to deliver,” said Ciovacco. The process started with demos and presentations, working backwards from the outcomes the university wanted to drive. The university also worked with AWS Partner ScaleCapacity to build a retrieval-augmented generation (RAG) pipeline that enables users to upload documents and incorporate them into AI conversations.
In early 2025, the university ran a trial to understand campus needs. By March 2025, the team launched development with an ambitious goal: to deliver the first version of boisestate.ai by the fall semester in September.
Building a multi-model platform on Amazon Bedrock
Supported by AWS, Boise State’s small team built the platform in just six months. Today, the web-based AI platform boisestate.ai serves the entire university through a single interface. It features a familiar chat experience with built-in AI literacy resources, such as example prompts and responsible-use guidance, designed to encourage good AI habits.
The platform also includes custom assistants—AI interfaces paired with specific instructions and uploaded documents. Faculty can create an assistant for their class by uploading syllabi and readings, and students can ask questions about assignments and due dates.
Amazon Bedrock provides the foundation for boisestate.ai, giving users access to multiple large language models, including Amazon Nova, Anthropic Claude, and Meta Llama. Users can choose the model that fits their needs, whether they need a larger context window, cost-effectiveness, or the latest capabilities. “It’s really nice to give our campus community the flexibility to choose any of the models they feel is appropriate for their discipline,” said Merrell.
Supporting this multi-model approach, the front end is hosted in Amazon Simple Storage Service (Amazon S3) buckets, while the back-end API runs in containers. Amazon Relational Database Service (Amazon RDS) for PostgreSQL handles user data, and Amazon DynamoDB stores conversations and messages. AWS Lambda functions power serverless tools that extend capabilities, including an integration with Semantic Scholar, a database of academic journals. To maintain compliance, AWS Control Tower enables centrally managed accounts that inherit security settings aligned with FERPA requirements.
Achieving 80% cost savings with organic growth
With AWS, Boise State targets three dollars per user per month, and current usage is under that target—more than 80% cost savings compared to the $7.2 million annual cost of individual commercial subscriptions. Instead of rushing an aggressive campus-wide promotion, the university took a soft launch approach in the first semester, prioritizing quality and user experience while gathering feedback before a broader rollout. Despite minimal marketing, adoption has grown organically month over month.
Faculty feedback has also been positive, even among those initially skeptical of AI in education. “I have to give kudos to folks on campus that are willing to engage with it, roll up their sleeves, and learn,” said Merrell. “We’re engaged in the hard, messy work of trying to figure that out as a campus.” The platform has also brought departments together in new ways. Instructional designers, for example, can now create and share assistants to help with tasks like building assignment rubrics, which faculty adapt for their courses.
Looking ahead, the university plans to integrate boisestate.ai more deeply with campus systems. Students will be able to ask about their academic advisor, search for courses that fit their schedule, and interact with their Canvas assignments—all through natural conversation with AI that understands their needs while maintaining privacy and security.
Lessons for other universities
With boisestate.ai now live and growing, Merrell offers several lessons for peers in higher education:
- Start with a narrow focus.
- Connect the project to return on investment.
- Check for existing campus AWS agreements.
- Know that small teams can succeed.
Boise State’s approach offers a roadmap for other institutions wrestling with AI adoption. The university built a platform that handles the core challenges—cost, compliance, and equity—while maintaining control over their data. Working with AWS, they created a system that gives everyone on campus access to AI tools without breaking the budget or compromising on security.
Learn how AWS helps education institutions innovate with generative AI.