Videos

Every EE project is site a knowledge-base wearing a website costume. Channels, fields, entries — You've already built the important part. This talk explores how to turn that structured content into an AI that actually understands your client's business. Retrieval-Augmented Generation ("RAG") enables AI systems to answer questions or perform tasks using YOUR content. This presentation will demystify RAG, in familiar EE terms — from the basics, all the way to more exotic concepts like embedding, vector search, and ranking. Then, we'll walk through setting up a minimal retrieval layer on a real EE site. We'll compare some different search/indexing methods (Pro Search, Dexter, custom embedding). And we'll wire up a simple query-answer flow, all using tools that fit comfortably into an existing EE stack.

You'll leave knowing exactly where RAG fits into an EE project, what it costs (in both dollars and dev time), and most importantly, how to start pitching it to clients right away.

The possibilities are endless:

• "Ask this site" — Replace the dusty old search bar with a chat UI that actually answers questions in the client's voice, sourced only from their real content (no generic ChatGPT vibes or hallucinations).

• "You might also like..." — Retrieval-powered recommendations that leverage deeper meaning, not just keywords or categories.

• FAQ generation — Automatically surface answers to common support questions by retrieving straight from product pages, docs, or blog archives.

• Research assistants — For content-heavy projects (law firms, universities, nonprofits) — Let site visitors query years of published articles, case studies, or policy docs, conversationally.

• Product catalog Q&A — E-commerce or catalog-driven sites get a natural-language assistant grounded in product entries and spec sheets. ("Do yall have that little red thingamabob that fits on my whatchamacallit?")

• Automated content audits — Find outdated, contradictory, or duplicate content across hundreds of entries; A killer upsell for your "we haven't touched this site since 2019" rescue client.

• Multi-language support, without a full editorial rebuild — An LLM can answer questions in a visitor's language, while pulling from a single-language content source.