A chat model writes tidy solver code for five products and falls over on a real catalog. ORPilot interviews the user first and keeps raw data out of the prompt, and that pipeline is worth copying for AI features on a WooCommerce store.
An AI pass over your products can finish with valid JSON and no errors while it quietly corrupts the data. A better prompt won’t fix that. The queue WooCommerce already ships will: one item per action, retries, checkpoints, and output you can audit with a script.
An AI feature that works for three turns and then dies with a context length exceeded error has a design flaw. The transcript was never a database. Here’s where the run log belongs and how to keep prompt caching working.
A customer asks you to delete their data, the erasure report says zero items, and your chat plugin still has every conversation stored. ERC research wants exactly those cross-session profiles, and WordPress privacy tools only see what you register.
WooCommerce 10.9 adds seven canonical product and order abilities and starts retiring the REST-shaped MCP beta. Most stores can wait. If you call /wp-json/woocommerce/mcp you have a migration to plan, and extension authors have two traps to watch for.
Advice on agent costs splits between free-roaming agents and tightly scripted ones. Exploring once and replaying the path makes sense, but in WordPress that replay is a cache, and the part that bites is invalidation.
A study tested ChatGPT on econometric code in Python, R and Stata, and it did well except in the language with the least public code. Legacy WordPress and WooCommerce work has the same weak spot, so I benchmark AI output before I trust it.
A recent piece argues most AI agents fail because teams build them top-down and hope the model holds it together. I agree, but WordPress has no long-running PHP process, so that orchestration has to live in hooks, the database and a queue.
A Towards Data Science piece builds one retrieval system four times, from TF-IDF to fine-tuned DistilBERT, and the fine-tuned model flips its answer between two epoch counts. For a WooCommerce store, that argues for pretrained embeddings and no training.
WordPress 7.0 Armstrong puts an AI layer into core: an AI Client, a Connectors API, and one screen for provider keys. I look at what it changes for plugin code, what stays off by default, and what to test before updating client sites.