Facebook, in its recent F8 conference shone the spotlight on the next emerging interaction model – bots. Also known as chatbots, this is something that’s likely to pick up steam going forward. Which begs the question – why now? Like in many aspects of technology, this prediction is based on the right set of environmental conditions. There is now almost a perfect storm of enabling factors, including:
- Improved artificial intelligence capabilities
- Strong integration with back end systems
- Availability of api’s within an increasing array of communication apps
Let’s look at each of the above in turn.
Possibly the biggest enabler of chat bots, or any other form of parsing natural language input to do various tasks on behalf of the user is the increasing maturity of artifical intelligence capabilities. Over the last couple of years, there have been rapid strides in this field – just witness the pervasiveness of personal assistants like Siri, Google Now or Microsoft’s Cortana. These capabilities are built upon core AI engines that are able to parse user inputs, piece together contextual information about the user, in order to do her bidding. This interaction model works for the “solved” problems that these digital assistants support. So, once the developers have programmed Siri to handle making appointments, take notes or create reminders, the software smarts, along with voice recognition are able to piece together the necessary instructions to carry out the task. But the flip side of this model is that it is limited to the capabilities that are already supported (see below for an example)
This limitation is due to the fact that the interaction relies completely on a synchronous communication model. This means that to preserve the user experience, the capabilities have to be rolled out carefully. Bots break this limitation by having the ability to be served by both system and human actors. After all, in a modern twist on the Turing Test, it doesn’t matter to the user whether the respondent is a software component running on a remote server or a human in a chat-enabled “call center”. This allows in-line interventions by humans (actual conversation or quality control) of the interaction. Potentially, this may allow more complex interactions to take place and be rolled out faster than the voice assistants currently prevalent. A good case is banking, where there are said to be more than a 1000 ways in which to query the account balance. Now, one way is to ensure that we have covered all these bases (and new ones as they emerge) in our application, the other way is to start when you get to 70-80%, and then supplement the boundary cases with human interaction. This is a form of “force multiplication” much like the IVR application does in a conventional, voice call center.
Strong integration with back end systems
For line of business applications, also known as systems of record, the ability to expose their data and capabilities on new types of channels has been one of the focus areas over the past decade or so. Enterprises have tried a variety of approaches and we have seen a number of “silver bullets” over the years. Starting from the days of EAI (Enterprise Application Integration) through to the big iron transaction management systems like TIBCO, a number of approaches and solution patterns have been tried, discarded and evolved. Big companies have spent a lot of money implementing and supporting this class of connective tissue, now also known by the generic term – middleware.These approaches have now matured to a point where there are now standard “recipes” for achieving these integrations – taking systems of record, some of which are based on technologies that predate the Internet and mobile eras, and exposing them to the Web and mobile channels. The way in which these integrations are achieved allows new and emerging form of channels (such as voice, chat, wearables, etc) to leverage the work already done to connect these back-end systems to the outside world.For organizations that have already done the hard work of doing these integrations, connecting newer channels such as bots is relatively simple. The difficult part is the AI and analytics, not the integration.
Availability of API’s
The old paradigm was that the browser is the universal platform. That has changed some with the advent of smartphones and app stores. The paradigm of downloading individual apps for specific tasks works in most cases and is still relevant, but if we have learned anything from the mobile trend, we have to provide users with choices for them to get their work done. One size does not fit all. If we have to go where the users are, we cannot ignore messaging apps. Today, messaging apps occupy increasing amounts of user mind share and their time spent online. It makes sense to engage the users where they do spend their time rather than make them change their habits.When seen in the context of messaging apps, there is also the social opportunity – the ability to leverage the user’s social graph in the interactions. While these social elements are not yet fully thought through, they will form a new and exciting frontier going forward.
With Facebook now getting aggressive about bots and with others like Telegram already implementing a bot platform, the front end capabilities to start the journey are taking shape.
These are early days and this new interaction model will continue to evolve. Customer acceptance is not assured – challenges include chat and messaging overload, the risk of spammy bots and an as yet unclear secure communications and payment channel between the ultimate service provider (like 1-800-FLOWERS) and the user. Also it remains to be seen whether users would be comfortable engaging a bank bot, for instance through a chat program.
These challenges aside, there is a lot of excitement surrounding this new frontier. There are some obvious advantages – no one wants to listen to the elevator music of a typical call center (in addition to being told repeatedly how important their call is). The bot model offers a way out of this. Ultimately the sheer convenience may just outweigh the adoption hurdles. Watch this space.