Develop a conversational search experience without rebuilding your app | ODSP922
Greg Crist demonstrates how to add LLM-powered conversational search to an existing production application using Elastic Agent Builder and a hosted model on Azure OpenAI, focusing on wiring retrieval into existing search workflows to deliver a production-ready conversational experience.
Overview
This on-demand Microsoft Build 2026 demo shows how to extend an existing application with a conversational search experience without rebuilding the app from scratch.
Key themes covered:
- Using Elastic Agent Builder to create an agent workflow that can be embedded into an existing production system.
- Integrating a hosted LLM on Azure OpenAI.
- Connecting retrieval to existing search workflows so the conversational experience can return grounded results.
Session structure (chapters)
- 0:00 — Microsoft AI ecosystem and Elastic data integration explained
- 02:02 — Introduction and agenda for the Elastic Agent Builder demo
- 10:02 — Starting to build the “Trip Planner” agent workflow
- 14:19 — Tool test validates returning the same customer details
- 14:35 — Creating an agent that uses the workflow tool
- 15:02 — Configuring the “Trip Planner Agent” (ID and instructions)
- 19:02 — Integrating agent functionality into existing apps
- 20:37 — Testing trip planner functionality inside the app
- 21:58 — Displaying trip plan output (weather, gear recommendations, itinerary)
What the demo builds
Trip Planner agent workflow
The session walks through building an agent workflow (a “Trip Planner” example) and validating it with a tool test.
Agent configuration
The presenter shows configuring a new agent (“Trip Planner Agent”) including:
- Agent identifier (ID)
- Agent instructions (prompt/instructions used to guide behavior)
App integration
The demo then focuses on integrating the agent into an existing application and testing the end-to-end experience, including displaying:
- Trip plan
- Weather
- Gear recommendations
- Itinerary
How Azure OpenAI fits in
Azure OpenAI is used as the hosted model provider for the conversational experience, enabling LLM-powered interactions while the retrieval/search wiring keeps responses connected to the application’s existing search workflows.