What is Beyond Apps? The Dawn Of AI Bots

Whether You like it or not - Messaging platforms and Natural Language Experiences are here to stay and they'll change traditional business models. Bots are taking over.

Few weeks back while I was flying to Bangalore from Dallas, American Airlines announced the cancellation of my flight (they do that for fun at times). I was asked to call their support desk number to change my tickets. Well, I explained my intent but I was politely requested to re-purpose my statements few times - and it took some time for me to realize that I was talking to a Bot. 

Of course, he (or she) was not the first smart Bot that I interacted with recently – I had similar experience with Bank of America and an Insurance agency’s Bot who called me for offering some advice. I’m not even mentioning the slew of bots I’ve configured for Slack, or Siri/ Cortana I use at times.

There are multiple options to build your own bots these days - ranging from a number of Bot back end startups, to biggies like Microsoft. Microsoft and Facebook have built/acquired their own AI back ends recently -  Luis and Wit.Ai– and they made these platforms easily accessible to developers with custom Software development kits. Both these companies were busy integrating their back-end AI platforms to multiple channels – For example, Microsoft Bot Framework can expose the bot you built to a multitude of channels out of the box including Skype, Text, Slack etc.  These platforms also give a set of abstractions for developers to define the meta model and train them too. Also, it is relatively very easy to put a voice layer on top of this.

“There is an App for that” is no longer relevant – there is a better way to reach your customers, and that is a Bot.

How Bots Work

Bots can listen to your conversation and respond to you. That sounds trivial, but this requires the Bot to do few things

1 - To understand the user’s intent.

For example, if you are building a Personal Assistant Bot user’s might say anything from “I want to Order a Pizza” to “Schedule an appointment with John Tomorrow”.  The first step is to classify the user’s utterance/sentence to a meaningful intent – in this case, let’s say OrderIntent or ScheduleIntent

2 - Extract the parameters/entities related to the context of the intent

Now, for each intent, we’ve to extract extra inputs. For example, if the intent is to order something, we should capture what the user wants to order, may be from where he needs to order it. Chances are that he may want to specify the details of what he want to order as well. Let’s say, the Bot should be able to translate the sentence “Schedule an appointment with John Tomorrow” to a function of type ScheduleAppointment(when, where, whom, ..,.. ).  The Bot can enter into a dialog with the user to capture more details – most Bot frameworks allow developers to build a dialogue based on a Form structure, much like how users fill a web form but more natural.

3 - Send back a meaningful response and/or take an action and continue to build up the context.

Bots that can understand user intention and translate the related inputs to real actions are changing the game for once and for all. With easy to configure-and-train Deep Learning and AI back ends, building specific purpose Bots are easier than ever. Ranging from Bots that can order pizzas and books for you, to Bots that can imagine and visualize, to Bots that can control your wearables and things – we are about to witness a Bot revolution.

What this means for your Business?

There are a ton of use cases where Bots will naturally fit in. Let us have a look at few scenarios where Bots could really establish a strong business case.

1 – Better market penetration

As messaging apps are the new platform, integration of Bots into these platforms can open up multiple channels for businesses and will enable penetration to a larger market. Lets talk about enhancing your product with a natural language bot that can be used to control it - like turning your room's light on or off, or a bot can do your plant's watering (hey Philips Hue and Parrot). Or, lets talk about bots that can order things for your customers.  “There is an App for that” is no longer relevant – there is a better way to reach your customers, and that is a Bot.

2 – Analytics and Predictions based on natural language

Who won’t love asking questions to get instant answers like “What will be our revenue for next quarter?” and “Which business unit performed the best in EMEA”?. Yes, you got it. Siri like experiences for enterprises are in the waiting.

3 -  Cost Reduction

Employing natural language based bots that can guide users in areas like customer support will give better customer satisfaction and cost reduction. Oh wait, isn't this what most of the banks and airlines already doing? Very soon, we are talking about Bots that can deal with customers in better ways by appeasing them and offering them goodies in case their sentiment is going south when they call in (of course after some fact check - let's say if the shipping is delayed for few days). Presently most of the automated voice systems are not up to their human counterparts – but these back-end systems will soon evolve and will get better in future to give a near-human experience, and they can be nicer than human customer service representatives for sure ;)

4 – Better customer service and support

Scaling software is easier than scaling humans at least in Europe and America. This will also make sure a minimum level of quality when it comes to areas like Customer Service to handle customer queries, give specific information like order status, account balance etc. And the next level is to help customers to do actions – like booking a ticket for a movie in the nearby theater or even for complex actions like Filing taxes where an interactive experience is required (listen to this H&R Block).

5 – ‘Guru’ Bots with special domain knowledge

Building Bots that can understand and respond to user intents specific to a domain has got a lot of potential. For example, soon we will see bots that are experts in areas like Baseball, Swimming etc – and conversational experiences with these Bots will engage people in various areas like Patient Care, Education, Diagnosis of problems in domains. Or even in areas like Quality control where you ensure quality by asking a series of questions and making decisions based on the answers.

How to Build These Bots

The easy path to build bots include using a platform like Luis or Wit.Ai – Normally this involves

1 - Setting up Intents, Parameter Types/Entities and training the system using sample dialogues (utterances) to detect the intents and parameters from utterances. The philosophy and abstractions will vary a bit from platform to platform.

2 – Building a conversation engine on top of this that can translate the detected intents and parameters to meaningful actions, and to send responses.

3- Providing integration with channels – For example, Microsoft Bot framework integrates with Slack, Skype, Text etc. And Facebook’s Bot engine can integrate to Facebook Messenger etc

Welcome to the era of Bots. It's already there in your favorite chat app or messaging platform, or on the other side of the phone when you call for customer support. Enjoy.