AI agents in practice - beyond the prototype
summary
Describes building agents that move from flashy prototype to production: a spreadsheet screener processes thousands of rows and improves decision quality while saving researcher time; a releaser automates four triggers to produce daily releases and cut merge conflicts; a backlogger and scripts trim token costs; a generalist called Grace the generalist investigates, writes PRs and self‑suggests improvements under human guardrails; model defaults center on Claude Sonnet/Opus/Haiku for tradeoffs in cost and capability.
AI-written summary · about
chapters (8)
- 0:00AI agents in practice
- 3:11From prototypes to useful agents
- 8:44Post-prototype blues and optimization
- 12:34Workflow agents as team members
- 27:13Level two agents and emergent behavior
- 37:40Safety trade-offs and governance
- 40:25Agent platforms and composability
- 44:38Key takeaways
related talks
40mAI Agents in Practice
07h 47mAI Agents Conference 2026 - Day 5
40mHow To Tame Your Agents?
30mAgents Building Agents
49mAI Agents: Collaboration in Real-Life Applications
36mBuilding Agentic AI Assistants: Lessons from productionising agents in Preclinical R&D
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
35mThe Agentic AI Engineer