What is genuinely new, what to test first, and how to prompt it

Practical guide — September 2026

The short version

GPT-6 Astra is not most interesting because it can write a slightly better product description or ad. Its real advantage is that it can take on a larger job: gather information, work across files and tools, make decisions, produce finished assets, check its work, and keep moving through a multi-step workflow.

For ecommerce teams, that changes the useful question from:

“What prompt should I use to generate a product page, email, or five ad variations?”

to:

“What complete workflow could I delegate—from inputs to a checked, usable result?”

The best mental model is:

• Astra is the brain: it interprets the goal, plans, reasons, and adapts across channels.
• ChatGPT Work or Codex is the workspace: it gives the model a place to work with projects, files, the browser, and longer-running tasks.
• Plugins, connected apps, and computer use are the hands: they let it retrieve information or act in tools.
• Skills and instructions are the playbook: they define how your company wants repeatable work done.
• You are still the accountable operator: you choose the objective, define boundaries, approve consequential actions, and judge whether the result is commercially sound.

That distinction matters. Astra cannot read Shopify, Amazon, Walmart, Meta Ads, Google Ads, Klaviyo, your CRM, Google Drive, Slack, or your inventory system unless the relevant information or access is available. A more capable model does not remove the need for clean data, permissions, checks, and human judgment.

Where Astra fits across the ecommerce business

This guide applies to DTC brands, marketplace sellers, omnichannel retailers, agencies, and B2B ecommerce teams. The most useful opportunities sit across the whole commercial system—not inside one marketplace.