The AI Commerce Gap: Why E-Commerce Operations Still Need Humans

Summary

Seven e-commerce executives and industry veterans argue that current AI commerce tools have not made large-scale store operations autonomous. Agencies and sellers still work across fragmented software: one tool detects a problem, another suggests a fix, and a human decides whether the recommendation is safe, applies it in systems such as Amazon Seller Central, and checks the result. They identify four major bottlenecks: catalog exceptions such as suppressed listings and compliance issues still require manual resolution; tools optimize isolated metrics without understanding budgets, inventory, margins, brand stage, and realistic goals together; feeding raw Amazon data to an LLM creates noise and unreliable recommendations because the model lacks operational context; and recommendations often do not flow into the systems where changes, support cases, and follow-up must be completed. The proposed AI-native system would maintain persistent merchant context, permissions, deterministic data preparation, and feedback loops. It would detect, diagnose, fix, verify, and escalate issues, using confidence-based automation: reversible low-risk changes could run automatically, while consequential catalog or compliance changes would require approval. The contributors expect AI agents to handle more repetitive seller operations over the next two to three years, while also influencing how shoppers discover products. They say product data and business rules will need to be structured so agents can understand and act on them safely.