Fortinet Reduces Cost and Latency for Its Customer Support Chatbot with Amazon Nova Micro

Fortinet’s comprehensive portfolio of cybersecurity products helps its customers—ranging from local school districts to some of the world’s largest enterprise and governmental organizations—get the protection they need. But its highly technical products and multiple versions of documentation caused its database to grow in size and complexity, impacting customer service. “As our product portfolio expanded, we faced the challenge of making our documentation accessible to more than 800,000 customers,” says Keith Choi, director of product management at Fortinet. Accurate, current documentation is vital to helping customers unlock the full value of Fortinet’s solutions and protect their organizations. But searching Fortinet’s documentation portal could be time-consuming for customers.

As an AWS Partner and avid user of artificial intelligence (AI), Fortinet saw great potential in using generative AI on AWS. Fortinet’s technologists determined that a generative AI–powered support chatbot would be an ideal initial use case that would address a critical customer issue and deliver high value. From quick access to much-needed documentation to higher customer satisfaction, the benefits of more efficient and effective product support were compelling.

However, this couldn’t be an ordinary chatbot. Given the importance of Fortinet’s cybersecurity products to its customers, the solution had to meet very high criteria. For example, it needed to provide accurate and relevant responses quickly, reducing the heavy lifting for developers and improving customer satisfaction. The company also wanted a solution that could scale across over 60 products and multiple versions and accommodate more products in the future. Other requirements included being cost-effective for large-scale deployment and working seamlessly with Fortinet’s documentation portal and content management system.

The bottom line was that the support chatbot had to help hundreds of thousands of customers, in addition to internal users, to effortlessly and rapidly find the exact information they needed.

To create its support chatbot, Fortinet worked with the Generative AI Innovation Center, which lets customers prioritize use cases, build road maps, and move solutions into production. Working closely alongside the center, Fortinet built a complete generative AI chatbot using Amazon Bedrock, a fully managed service that offers a choice of high-performing foundation models. The Fortinet team explored multiple models during the evaluation period.

“We heard remarkable figures regarding the performance of Amazon Nova models at AWS re:Invent 2024,” says Choi. “When experimenting with Amazon Nova, we found that not only was the quality high, but its pricing was very attractive.” Because its generative AI support chatbot would be used by an extensive customer base, Fortinet gave priority to cost efficiency. So it chose Amazon Nova Micro, which could offer 85 times reduction in inference costs compared with its previous model. “We already had the chatbot’s architecture up and running,” says Choi. “But Amazon Nova Micro was the sweet spot we were looking for, and Amazon Bedrock gave us the flexibility to switch to Amazon Nova Micro quickly and easily.”

Fortinet’s support chatbot benefits from the capabilities of Amazon Bedrock. For example, using Amazon Bedrock Knowledge Bases, the company can give foundation models and agents contextual information from its private data sources to deliver more relevant, accurate, and customized responses. Also, Amazon Bedrock Guardrails lets Fortinet implement safeguards that are customized to its application requirements and responsible AI policies.

Fortinet is planning widespread public use of its generative AI–powered chatbot. The company can now filter queries, separating simple requests from those that require deeper knowledge and attention from specialists. Thus, its team has more time to focus on the core mission of delivering a complete and fully immersive user experience. “Our innovative generative AI–powered chatbot is transforming how we support and communicate with our customers by delivering trusted and relevant best practices,” says Robert May, executive vice president of technology and product management at Fortinet.

Fortinet’s marketing team sees the chatbot as proof of the company’s ongoing commitment to using cutting-edge technology to serve its customers. “Anytime we can streamline the experience for our customers—any touch that they have with the company—it’s a win for everyone,” says Mishel Fletcher, director of cloud alliance marketing at Fortinet. “The new chatbot is a big step up for us, one that gives us a competitive advantage.”

Looking to the future, Fortinet is already exploring new use cases for employing Amazon Nova models in customer support and other key areas of its business. “By working alongside AWS, we get enterprise-grade security and scalability,” says Choi. “The ability of Amazon Nova Micro to scale while maintaining low latency is important as we move from an internal audience to hundreds of thousands of customers. And by using Amazon Bedrock Guardrails, we get greater control over the user experience and behavior.”