Scaling AI-powered security training using AWS with KnowBe4

Scaling security training and developing AI-powered capabilities requires an architecture that can respond to changing demand and connect data across products. By building an event-driven architecture using Amazon Web Services (AWS), AWS Partner KnowBe4 addressed these needs and created a data foundation for its AI-powered security capabilities. The company combined this foundation with common AI development patterns and a culture of continuous learning. Small engineering teams supported by AI coding agents now run dozens of AI projects, with KnowBe4 reporting productivity on net-new development of up to 10 times that of 12–24 months earlier. Its customer-facing AI agents also automate phishing simulations and training.

KnowBe4’s security awareness training product experienced traffic surges as administrators assigned training at the start of the week, month, and business day. Demand followed local business hours across the United States, Europe, and Asia Pacific and Japan. Before provisioning infrastructure for peak usage, the company experienced processing delays and backlogs during busy periods. Capacity constraints had also caused product outages.

The monolithic application behind the company’s flagship training product made scaling individual workloads difficult. As KnowBe4 expanded its portfolio and acquired companies, sharing data across products required repeating point-to-point integrations. That growing complexity created a need for a common way to exchange data. The challenge also extended to AI development, where the company needed well-structured product data to build AI-powered security capabilities. To support changing demand and simplify data sharing across products, KnowBe4 chose to modernize its architecture using AWS.

KnowBe4’s infrastructure, DevOps, and site reliability engineering teams chose AWS based on their experience using its services and its pace of innovation. Engineers defined the requirements and success criteria for a global event bus, and then worked alongside AWS experts to design and scale it using Amazon EventBridge to build event-driven applications at scale. KnowBe4 built this event-driven architecture in about 8 months, creating a system internally called the Data Foundation Service. “It gave us the event-driven backbone our AI strategy runs on. By decoupling our systems, we can feed clean, real-time product data to our AI agents and models, and ship new agentic capabilities faster, with an isolated scope of damage if something fails,” says Andy Suarez, senior vice president of engineering at KnowBe4.

That event bus now supplies data to the data lake supporting KnowBe4’s AI and agentic features. AWS provides built-in security capabilities and guidance to help the company develop and scale applications securely. KnowBe4 stores data lake content and application data in Amazon Simple Storage Service (Amazon S3), object storage built to retrieve any amount of data from anywhere, and it runs Apache Spark–based extract, transform, and load workflows using Amazon EMR, a big data processing service, to transform data in its lake.

For AI development, KnowBe4 uses Amazon Bedrock for building generative AI applications and agents at production scale. The company accesses Anthropic Claude models, as well as Amazon Nova models, which are fast and cost-effective foundation models for text and multimodal needs. KnowBe4 also uses Amazon SageMaker, which provides an integrated experience for analytics and AI, to run open source models, including an embedding-based agent. Keeping AI development within AWS helps KnowBe4’s engineers get started quickly and use common development patterns across model providers. The company also reinforces continuous learning through an internal certification program with incentives for engineers to earn associate-level certification within their first 12 months.

The Data Foundation Service shipped in about 8 months. Within several hours of implementation, it deployed two features to production 3 months ahead of schedule. Building on that foundation, the company now has dozens of AI projects in development, run by teams of two to four engineers supported by AI coding agents. KnowBe4 reports engineering productivity on net-new development of up to 10 times that of 12–24 months earlier.

KnowBe4 used its AI foundation on AWS to build AIDA Orchestration, an AI agent that automates phishing simulations and security awareness training and is included in KnowBe4 SAT Advanced, the company’s advanced security awareness training tier. “Building on AWS helps us ship new risk indicators faster. We're at 316 today, and the number keeps growing every year,” says Matt Duren, senior vice president of engineering for AI and data at KnowBe4.

Based in the United States, KnowBe4 is a market-leading cybersecurity company focused on securing the whole workforce.