Building Reliable Browser-Use Agents for Storefront Optimization
summary
Autonomous browser agents shop real Shopify storefronts at scale, using a perceive plan execute update loop and four engineering pillars: episodic memory, multi-modal perception, scalable infra, and systematic evaluation. The platform runs up to 24,000 sessions concurrently (600 agents per storefront scenario), uses session memory and vision to boost success rates by ~30–40% with memory and +5–10% with vision, achieves 90% vs 45% success with/without memory, and predicts conversion direction ~70% of the time in under an hour.
AI-written summary · about
chapters (12)
- 0:00Browser-use agents for e-commerce
- 1:44Browser agents for storefront optimization
- 4:11Heterogeneous storefronts & agent architecture
- 5:16Episodic memory for multi-step tasks
- 7:44Episodic memory accumulation
- 8:06Multi-modal web page perception
- 8:50Scalable infrastructure for parallel sessions
- 10:16Evaluating agent reliability
- 12:46Real-world validation results
- 13:07Simulated A/B testing & storefront health
- 14:09Four pillars recap
- 14:51Q&A
related talks
30mAutonomous Browsing using Large Action Models
37mLeveraging Generative AI for Enhanced E-commerce
21mThe Current State of Browser Agents
31mShopify's Ray-Powered Approach to Multimodal AI in E-commerce
01h 07mMastering Complexity and driving Growth
43mThe AgentVerse - Build fast, Scale smarter
29mClient side Web AI Agents Building smarter user experiences for a future age
31mBuilding Powerful Marketing Apps