Predictive bots

This post is going to be a short follow-up to last week’s post on bots. In the earlier post, I basically shared the following thoughts:

  1. For bots to be widely used, it needs to be easier for people to perform an action by interacting with a bot than to take the action themselves in an app or on a website.
  2. Among the four use cases currently possible for bots, which include alerts, search/inputs, support, and bookings, I believe that the first point only holds true for support.
  3. The reason why I see support as the most suitable use case for bots is because it requires two-way communication. The other actions are examples of one-way communication and are therefore easier for someone to execute themselves rather than by interacting with a bot. For two-way communication, however, interacting with a bot can be more accurate and faster than interacting with a human agent.

The underlying assumption behind these thoughts is that a human is initiating each action. If a human decides to receive an alert on a particular topic, perform a particular search, or make a particular booking, it’s very effective for the human to do so themselves. They don’t need a bot.

However, what if an intelligent bot were the one initiating the action? By taking into account your profile, past behavior, and contextual information, an intelligent bot could alert you on a topic that you didn’t know you would be interested in, share the results of a relevant search that you would eventually be performing before you identified the need to perform it, or recommend that you buy a good or service that you didn’t even know you would enjoy.

In each of these cases, an intelligent bot could use its predictive power to push you to take an action before you decided to do so yourself. This effectively transforms what were once examples of one-way communication in use cases like alerts, search, and bookings into examples of two-way communication. And as established in my earlier post, bots can be very useful for two-way communication.

So, to summarize, in the absence of predictive intelligence, I think that support is the one use case where bots are likely to be widely adopted. But, together with predictive capabilities, intelligent bots could gain widespread usage across virtually any category including alerts, search, and bookings. The greater their predictive capability, the more we’ll use them.


Also published on Medium.