On April 12, 2016, at its F8 developer conference, Facebook announced the Messenger Platform in beta and opened it to bots. The pitch was that conversation would become a primary computing interface - that instead of installing apps, people would simply message a business. Facebook’s own announcement described what the bots would do: “automated subscription content like weather and traffic updates, to customized communications like receipts, shipping notifications, and live automated messages all by interacting directly with the people who want to get them,” delivered through a new Send/Receive API.
The industry took the cue. “Bots are the new apps” became a slogan; companies rushed to build conversational interfaces; commentators predicted the app store would give way to a chat window. Facebook had been priming this earlier with M, an assistant launched in 2015 inside Messenger that combined AI with human operators behind the scenes.
The wave broke quickly. Most bots of that era could not actually hold a useful conversation - they were brittle decision trees that frustrated users who expected something smarter. M, the flagship assistant, never expanded beyond a small test group and Facebook wound it down in early 2018. The grand interface shift did not arrive.
What makes this hype cycle interesting is that the underlying intuition was not wrong - it was early. The conversational interface did eventually become a serious computing paradigm, but only when large language models in 2022 and after made bots that could genuinely converse. The 2016 wave had the right idea and the wrong technology, and it collapsed in the gap between them.
Note: Facebook’s wind-down of the M assistant in January 2018 was communicated to users and confirmed in company statements relayed through the press rather than in a standalone first-party document; the bot-platform launch above is the first-party anchor for this entry.