From Blank Page to Prod: A Year of Building AI Agents at Carrefour
summary
Company deploys agents in early 2025 using Gemini and Google’s Agent Development Kit, integrating RAG, observability and an agent-analytics pipeline that streams events into BigQuery. A production agent reduces a two-day store-turnover assessment to 10 minutes vs 2 days by relying on structured SQL data, but integration into legacy systems and Google Chat costs 50–70% of engineering time. Concrete deliverables include session metrics, cost/latency tracking, evaluation tables and a working Chat-integrated agent. Agent Development Kit / Gemini / Google Chat API
AI-written summary · about
chapters (8)
- 0:00AI agent failures at Carrefour
- 1:28Why AI projects fail
- 2:33LLM overestimation for structured data
- 3:40Interactive docs with AI agents
- 4:53Boosting agent adoption
- 6:19Legacy system costs
- 7:11Monitoring agent behavior
- 8:59Year of AI agent lessons
related talks
45mCreate and deploy your first agent with Google Cloud Gemini Enterprise
01h 48mBuilding Conversational Agents
13mBuilding AI Agents is easy
01h 45mAI Engineering with the Google Gemini 2.5 Model Family
33mAI in action: Boost productivity and deliver business value with practical tools
45mThe Role of AI in Modern Marketing
49mLet's build an AI agent
31mBuilding, Deploying and Managing AI Agents at Scale