Engineering Production-Ready Onboarding Agents
summary
Describes building a production onboarding agent with an internal Python Agents SDK, Snowflake/Cortex-backed retrieval, and Temporal workers for orchestration; implements retrieval modes: exploration vs deep-dive, strict tool contracts for actions, streaming support and Python tool-strategies to force required calls. Tools include help-center retrieval, invoice and contract generation tools with prompt validation and UI integration, plus conversation-management tools (help-me-answer, notes, wrap-up agent). Deployed to 90% of trials vs 10% control and shows a 36% conversion uplift, so it ships and works.
AI-written summary · about
chapters (7)
- 0:00Introduction to onboarding agents
- 5:27Core components of the agent
- 10:03Retrieval tools and dynamic responses
- 15:19Generating contracts and invoices
- 18:58Production infrastructure with Temporal
- 22:53Measuring impact on trial users
- 23:57Key lessons and summary
related talks
32mThe AI Agent Path to Prod: Building for Reliability
07h 37mAI Agents Conference 2026 - Day 3
52mDesigning Production-Ready Agentic Systems: Lessons from the Field
18mFrom POC to Profit: Deploying Customer-Facing AI Agents That Actually Work
51mActually shipping agents and not just talking about it
33mAgent Experience
45mAI agents in practice - beyond the prototype
36mBuilding Employee Experience Agents with Microsoft Copilot Studio and MCP Integration