Chatbots are becoming as common as 1-800 customer service lines or website FAQs. You can say this out loud if it helps you get used to interacting with the automated assistants designed to answer your every question and command.

Designed to help guide us through both simple and complex procedural tasks inside the mobile, desktop, online, and even back-office network systems that we need to interact with, chatbots are essentially software applications.

Meet Three Types of Chatbots

Some chatbots are pretty basic. These chunks of code go to work in jeans and sneakers. They help users react to comparatively simple form-based applications (mostly web-based) and provide us with preprogrammed information, not dissimilar to what we might be able to find ourselves on a Frequently Asked Questions (FAQ) page of a website.

The “Smart” Bots

Other chatbots are more sophisticated. These chunks of code go to work in a business suit. They are built with more lines of code and boast more connection points to external information streams and resources, such as data analytics capabilities like those in SAP HANA.

The most advanced chatbots today are cultured and refined because they exhibit a real appreciation for the way we humans expect to interact. These chunks of code go to work in a tuxedo or ballgown. They are built to understand the contextual significance of the interactions they engage in.

Along with their capabilities for analytics, these high-end chatbots also have a more-sophisticated appreciation of the idiomatic nuances of human language and may be able to communicate using Natural Language Understanding (NLU).

Handoffs = Bot Learning Opportunities

Common to almost all chatbots will be a “handoff” gateway. This is an algorithmically defined cutoff point, capable of detecting when the human user’s requests outstrip any of the “currently known” capabilities of the chatbot.

That currently known quotient is important. A smart (tuxedo or ballgown) chatbot will be built with enough machine-learning-fuelled artificial intelligence (AI) to be able to fill gaps in its knowledge based on its mistakes.

To label them mistakes is actually a bit unfair to the chatbots. Let’s call them “failings.” To put it another way, smart chatbots will be able to learn what they didn’t know from the human support that fulfils the handoff request, and then capture that know-how in order to deliver that information automatically when responding to future inquiries.

How to Create the Perfect Chatbot

So, do we need to hand craft hard-coded chatbots for every scenario? In some cases, they will need extensive custom programming. But in many cases, customers can take advantage of structures and procedures (both human conversational and programmatic) that have been laid down as template-based best practices.

Hey SAP, Build Me a Chatbot

This is precisely what’s behind the SAP Conversational AI offering. Formerly known as Recast.AI, a company that SAP acquired in early 2018, the SAP chatbot platform is intended to offer a faster route to building bots with AI. Many of these can be constructed from the get-go as end-to-end products that use preconfigured bot technologies, all aligned to help improve the customer experience.

SAP explains that its SAP Conversational AI technology matches the requirements of conversational chatbots and includes high-performance Natural Language Processing that supports more than 20 languages. According to an SAP press statement, “SAP aims to simplify complex business interactions and processes by employing conversational user experience technology: The intention is to have applications speak to SAP software users in natural language.”

Prior to acquiring Recast.AI, SAP had already created the SAP CoPilot digital assistant, a web application and platform that dovetails with (and from) SAP Leonardo to build conversational applications. Going forward, SAP clearly intended to bring all conversational AI competencies together in one technology offering, again driven by SAP’s design and implementation insights.

So Human, and Yet Not Human

By 2020, Gartner has estimated that customers will manage 85 percent of their relationship with the enterprise without interacting with a human. Why are bots becoming so prevalent? One answer could be that people have become so comfortable interacting with their peers on social networks that messaging platforms are the next logical step: They are learning to like chatting with more immediacy.

How should we build the best bots for the future? Part of the answer comes down to us, the humans, being able to write very clear questions in the most direct language possible as we seek to train our bots to handle more complex future tasks. We need to think about the semantics behind user intent and be able to splice between civilities and frivolities as we cut to the chase and work out the true meaning behind each user request.

And those user requests will always reflect the eccentricities of the people who make them. Here’s just a small example. The app for the 2018 SAPPHIRE NOW and ASUG Annual Conference had a little extra help from some human monitors during the week of the event. Those monitors included representatives from ASUG’s IT team, who tracked and tagged the many questions attendees submitted to the chatbot. Some of the stranger ones? Attendees asked questions such as:

  • Where can I get a cocktail at 10 a.m.?
  • How can I find a date in Orlando?
  • Are you cute?

Helping Bots Crawl, Walk, and Then Run

Step one is a small one. We should break down topics into components and think about how we will structure bots to serve smaller tasks—jobs based around what we could call microcontent. Then we can aim to build more fully functioning, grown-up bots that can adapt to the complexities of human language and human nature.

Curious how you can bring artificial intelligence to your customer support processes? Watch our recorded webcasts on SAP Leonardo.

If you're looking for innovative ways to delight your customers, you can also join us at our Customer Experience Innovation Forum, our newest event for retailers, consumer products companies, and wholesalers.