The Big Bot Wave

Sunkara Swetha

Although artificial intelligence (AI) has been around for decades, of late a new wave of “.ai” is sweeping the technology world. This time it has manifested in the form of bots.

Tech giants of Silicon Valley have ventured into this space by building bots themselves — such as Cortana, Siri, or Alexa — and/or by coming up with bot-building platforms like Wit.ai or API.ai. Other smaller companies and startups have also been quick to jump on this bandwagon, resulting in an overflowing “bot landscape” in the otherwise human-dominated world.

But that’s just the tip of the iceberg! On Facebook Messenger alone there have been approximately 35,000 bots built in last six months — news bots, weather bots, and scheduling bots, to name just a few.

The need to appeal to millennials, who grew up with texting as a primary mode of communication, and the popularity of messaging channels such as Facebook Messenger, SMS, WhatsApp, and Slack, have paved way for the emergence of chatbots. Chatbots allow users to simply text their needs and receive a response. Having a personal connection with their customers in the channel of their choice has created a huge business need for chatbots in the enterprise sector. And the DIY, no-coding-needed bot platforms and several custom bot development startups are trying to fulfill the admonition: “Be where your customers are.”

The business need

Some say bots are the apps of the future, while others disagree. However, the important question to ask at this point is “Are these bots solving any core business pain points in the best possible way?”

Sadly, the answer is more no than yes. Critics say that the current bots are just bot-ified versions of existing apps, or fun toys at best. The bots that address this issue effectively, just like apps or any other software solution, are the ones that will succeed.

The best use cases for bots are the ones where there is a real need for communicating with a human, and therefore a preference for getting things done in natural language, or at least in a natural “flow.” It is essential to first carve out a tightly defined use case that fits this criteria, rather than trying to force-fit an existing app into a bot. Bot design —which involves building conversational flowcharts (or equivalent representations) of typical business flows and understanding what things are best done via rich UI elements vs. plain text — comes later.

Are banks ready for bots?

My decade-long career in the banking sector naturally makes me curious about how bots are transforming this industry. A super-intelligent, highly personalized bot that manages your finances, credit card expenses, mortgage payments, tuition fees, medical bills, and savings while simultaneously optimizing your investment portfolio based on market trends may sound like a fantasy.

Yet many fintech startups and some future-looking banks have started scratching the surface of the botiverse. One such startup is Kasisto, whose consumer banking bot MyKAI provides a natural language interface to all your bank and credit card accounts. You can ask the bot about your expenditures in various categories (food, entertainment, etc.) or even ask it to send money to your friends via Venmo. There are also several other personal finance bots available now, like Trim, Abe, and Penny.

As recent articles suggest, banks are also looking for new and better ways to communicate and serve their customers. For example, the Amex bot enables you to connect your Amex account to Facebook Messenger to receive notifications related to purchases, card benefits, promotions, etc.

Although these are nice use cases, they do not solve one of the most critical pain points that banks have: customer service. For example, to interact with your bank to make a credit card or loan payment, you still need to go via one of the traditional routes – app, website, interactive voice response (IVR), or a call to the customer support number. IVR systems are clunky and impersonal, and users often get lost in their complex labyrinth of button-navigated options. Phone calls for such mundane tasks are irritating for customers due to long wait times, and they can be quite expensive for the businesses to support.

Although apps and websites provide an asynchronous channel for customers, they can also be expensive to build, maintain, and customize across various platforms (iOS vs. Android vs. Windows), browsers, and form factors. Also, with the ever-increasing number of apps, the precious real estate of your phone starts getting cluttered and diminished.

A bot simplifies this. Bots seamlessly integrate within existing popular channels of communications using simple APIs, providing the “build-once-deploy-everywhere” advantage. The UI is relatively simple with a few basic components like choice menus, carousels, and URLs, making them easily adaptable for various organizations in an industry. The main challenge in bot-building is conversation design and understanding and responding in natural language, which in general is a significant challenge in the areas of natural language processing (NLP), AI, and machine learning.

So how do you get around this challenge? The key is to first identify common yet simple use cases that can be solved by a bot in a step-by-step fashion. For example, the most commonly performed operations during a loan’s lifecycle are generating monthly statements, installment/lump-sum payments, loan payoffs, or requests for extensions. Each of these tasks can be clearly explained using a simple step-by-step process without much customization (imagine the steps involved in paying your credit card bill).

So the first version of the bot can be purely menu-driven, to address these frequent use cases. More complicated processes like loan origination, refinancing, or renewal, which often need human involvement, can be tackled later. Also, the bot can be made smarter progressively by adding simple layers of NLP, for example, understanding dates/times, or simple intents entered via free text.

All in all, 2016 can be described as the year of bot emergence, and recent trends show that we can expect considerably more involvement of bots in our daily lives going forward.  Seventy percent of American consumers say they prefer using chatbots to interact with companies for common interactions over other channels. So bots and companies that cater to such needs will gain and flourish.

For more on how AI is affecting business today, see How Machine Learning Enables The Intelligent Enterprise.


Sunkara Swetha

About Sunkara Swetha

Sunkara Swetha is Principal Consultant for the North America region at SAP.