How will users discover bots when there are thousands of them? App stores just don’t scale. Google can be the first to find a new model of bot discovery. Brad Abrams, group product management of Google Assistant, touched the topic of bot discovery in an episode of O’Reilly Bots Podcast.
Google Assistant supports two models of connecting users to bots. The first model is “talk to” model. You call the name of the bot explicitly. This approach works fine if you know the brand. It is similar to an app store model.
The problem is that you may not know a brand. Then you can just tell the assistant what you want to do and it will choose the bot for you. An example Brad was talking about: “OK Google, I want to meditate,” and Google Assistant connects you to Headspace bot. Developers of Google Assistant Actions can register invocation triggers, which work as discovery phrases. Over the long term, Brad expects that most bot invocations will be done using discovery phrases.
Multiple developers can register the same discovery phrase. Ultimately Google will be able to do ranking on discovery phrases. When a user requests something, Google Assistant will use ranking signals to figure out which bot to recommend.
When it comes to search and ranking, Google has tremendous experience. Google can collect a lot of metrics to inform ranking: is someone is using the bot, are they repeat users, how many turns of the dialog do they have with it. You can also associate a bot with the website, and Google knows how the website would rank for a similar query.
When users look for an action, Google Assistant gives only one result. In the future, Google will try to give two or three results, and it will help Google Assistant learn which options users prefer when there is a choice.
This model looks very different from how Amazon Echo works, and it may be a model of bot discovery which will work well for end users as well as bot developers.