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.