Facebook testing Moneypenny, a personal assistant inside Messenger that consults real people for help researching and ordering products, among other tasks
Facebook is working on a digital assistant named Moneypenny to help you find and buy products — Facebook Messenger is getting …
Context & Ripple Effects
This report is the earliest sighting of what became Facebook's Messenger assistant program: Moneypenny pairs software with real people who research and order products on a user's behalf, making commerce a chat rather than a shopping trip. Weeks later, Facebook confirmed the product publicly as M, an AI-based assistant supervised by humans — same architecture, new name.
The arc since has been one of steady widening: opening M to all US users with money transfers and planning features, then positioning the assistant and AI chatbots as the conduit for bringing businesses and ads into Messenger. Moneypenny matters because it shows Facebook chose a hybrid human-plus-AI design from day one, not a pure bot.
First-order effects
- Test users can delegate product research and ordering to Moneypenny's human operators inside Messenger, putting purchase intent directly into Facebook's own conversation surface.
- The human-in-the-loop design sets the quality bar for the assistant Facebook would later ship as M, since trained staff absorb tasks the software cannot yet complete.
Second-order effects
- Merchants gain a new conversational channel: if the assistant intermediates orders, pricing and placement in Messenger become merchandising decisions, which is exactly the businesses-and-ads role Facebook later planned for AI-enabled chatbots.
- Third-party services follow the traffic — Pinterest's later Messenger bot and collaborative search chat extension show platforms racing to embed their discovery experiences where the assistant sits.
Third-order effects
- If assistants become the front door to purchases, messaging platforms shift monetization from ads adjacent to conversation to transaction fees and sponsored actions inside it.
- Human-supervised assistance looks like a transitional structure: humans seed training data and handle edge cases while automation scales, a template other platforms building embedded agents would replicate.
The trend: Messaging apps are evolving from communication utilities into assistant-mediated commerce surfaces, with the hybrid human-plus-AI assistant acting as the workflow front door for search, ordering, and payments.