Meta has opened an internal test of a personal AI agent, code-named Hatch, to some of its employees. Unlike a chatbot that answers questions, the agent carries out a series of concrete actions on the user’s behalf: filling out a form, booking an appointment, comparing products or making a purchase. The user must approve each action before it runs.
Meta trains the agent on simulated copies of real sites
To train Hatch before any public launch, Meta built simulated environments that replicate real services: DoorDash for food ordering, Etsy for handmade goods, Reddit and Yelp for information searches, and Microsoft Outlook for email. There the agent learns to navigate, fill in fields and complete a task without a human stepping in at each stage.
Instagram is the first shopping target
Beyond administrative chores, Meta is aiming at a use closer to payments. Users would tap a product they spot in a Reel or a post, and the agent would compare the options and complete the purchase without the user leaving the app. Meta’s stated goal is to take friction out of mobile shopping and compete with TikTok Shop, which already has a foothold there.
Rivals already close the sale inside the chatbot
Hatch enters a crowded field. OpenAI has offered Instant Checkout for several months, letting users complete purchases without leaving ChatGPT. Google is building agentic checkout paired with inventory checks at merchants, and Perplexity has relaunched its own shopping tool with more personalization. Meta is also running a separate, already public test that is distinct from Hatch: a carousel of product recommendations inside the Meta AI chatbot, shown to select US users. It displays an image, the brand, the price and a link to the merchant’s site, with no built-in checkout.
| Company | Feature | Built-in checkout |
|---|---|---|
| OpenAI | Instant Checkout in ChatGPT | Yes |
| Agentic checkout with inventory checks | Yes | |
| Meta | Meta AI chatbot carousel (test under way) | No, links out to the merchant |
| Meta | Hatch (internal employee test) | Planned, not yet public |
The big platforms’ ambitions far outpace actual use: just 2% of US digital buyers start with an AI assistant, according to EMARKETER. That makes Hatch a bet on the future rather than a catch-up move. The contest is not yet about volume. It is about delivering an experience reliable enough, in recommendations, product availability and completing the transaction, that users actually hand over the purchase instead of going back to clicking through it themselves.