Decagon Product Foundations

An AI agent is only as effective as the information it has access to. You can equip your Decagon agent with the knowledge it needs to answer questions accurately and consistently.

When customers ask for information, the agent responds using retrieval-augmented generation (RAG), pulling from your connected knowledge sources. When customers need to take action, like updating an address or disputing a charge, that’s when Agent Operating Procedures (AOPs) come into play. This module focuses on the knowledge side of that split.

The first step is connecting your existing sources of truth. Decagon integrates with help centers, internal docs, and ticketing systems, automatically ingesting updates so your AI always reflects the latest information. 

For content that isn’t public-facing, you can add Snippets — short, internal knowledge pieces that supplement your knowledge base. Snippets are useful for new product launches, promotions, or known issues that don’t belong in public documentation. They’re easy to manage and ensure the AI can respond with the right context without exposing sensitive details.

Over time, your AI can help improve your content. The Suggestions feature scans actual conversations to identify gaps where the AI agent lacked sufficient context and had to escalate to humans. These gaps are grouped by category, ranked by frequency, and in many cases paired with draft help articles that you can review. Suggestions run monthly by default, but can also be triggered on demand or exported for deeper reporting.

By combining connected knowledge sources, Snippets, and Suggestions, your Decagon agent becomes context-aware, continuously improving, and always grounded in accurate information.