Technology Blogs by SAP
Learn how to extend and personalize SAP applications. Follow the SAP technology blog for insights into SAP BTP, ABAP, SAP Analytics Cloud, SAP HANA, and more.
cancel
Showing results for 
Search instead for 
Did you mean: 
As the world becomes increasingly conversational, companies need to act fast, because this major shift impacts their two main audiences: employees and customers.

Nowadays, the majority of the enterprise software solutions on the market are extremely complex. As a result, they are difficult for employees to use in their everyday work without the assistance of an IT help desk. This often results in a poor employee experience – frustrating employees, lowering productivity, and tarnishing the company’s image as an employer.

Customers, on the other hand, are often exasperated because customer services are maxed out by the high number of inquiries. What’s more, Web sites are now more complex than ever, so people struggle to find the information they need to do straightforward things like ordering a new SIM card.

In both cases, chatbots provide the most viable solution when it comes to automating simple tasks and inquiries from customers and employees. We often underestimate the capacity of chatbots to improve with use, to deliver 24/7 assistance, and to create immediate value, inside and outside normal business hours and via any device.

But building a great chatbot is hard! In this article, we’ll explain how to build a chatbot that understands everything users with what we call the horizontal coverage. Let’s get started!

Note: This article assumes that you are familiar with the fundamentals of bot building. If you haven’t built a bot yet, check out this tutorial to learn how to build your first chatbot with SAP Conversational AI. You’ll also find a wealth of useful information in our product documentation.



The importance of designing your chatbot horizontally


In the design process, businesses need to ensure a user experience that leaves customers and employees satisfied every time they engage with a chatbot. Building a chatbot that understands only the five most frequent topics asked would work in theory. But in practice, users will ask anything but the planned use cases. This leads the chatbot to reply “Sorry, I don’t understand” too frequently, resulting in a poor user experience.

That’s why we decided to change our bot-building methodology and adopt a horizontal approach.

Building a chatbot horizontally means building the bot to understand every request. And this means building a horizontal coverage – in other words, a dataset capable of understanding all questions entered by employees or customers.

Creating a great horizontal coverage doesn’t necessarily mean that the chatbot can handle every request. However, it does mean that any request will be understood and given an appropriate response that is not “Sorry I don’t understand”.

With our methodology, once the bot understands the question, it can take one of the following actions:

  1. Handle the request autonomously with a fully automated conversation

  2. Guide the user to the right webpage to help them get the information they’re looking for


Of course, other actions are possible, depending on your industry or business flow. For example, redirecting to human agents is just one of many options.


The horizontal coverage concept




Build a smarter chatbot by integrating horizontal coverage


The key to building effective horizontal coverage is to efficiently collect conversation logs and feedback from your users. Let’s look at how to build a robust dataset by completing the following steps:

a) Online surveys


Surveys are a great way to gather user data, and user data is the core of powerful horizontal coverage. Start by creating an online survey to ask your potential users how they will interact with your product. The goal here is to enable you to think as if you were a customer who uses your product or an employee of your company:

  • What use cases are important and need to be covered by the chatbot?

  • What questions are users most likely to ask the chatbot?

  • What are the main pain points encountered by users of your product?


You will be able to collect valuable insights into queries made by your users, which will help you to identify strategic intents for your chatbot. It’s got to have at least a few thousand user sentences to build a great horizontal coverage. You can go up to 50,000 or more! Here are some examples of questions you can put to participants in your survey:

  • What would you ask an SAP SuccessFactors chatbot? For example: see my training plan; open a new position; update my personal info; see open positions; take a day off; create a spot award…

  • What are your biggest pain points when using SAP SuccessFactors?

  • Do you have a specific use case in mind where a human agent can be avoided?


b) Intent clustering


Once you’ve collected feedback from your users, you’ll be able to identify the following different types of intents:

  • New intents that have not yet been integrated into your bot training: these need to be created on the SAP Conversational AI platform and enriched with expressions.

  • Intents that have already been integrated into your bot training: These need to be enriched with new expressions.

  • Intents not related to the scope of your chatbot: These give you the opportunity to maximize your chatbot’s understanding by covering other use cases, enabling the bot to propose alternative solutions for the end-user.




List of expressions in your intent




This is the point at which you should work on your intent clustering. Use expressions mentioned by your users to enrich both new and existing intents so that you have as many expressions as possible.

c) Train your chatbot


Work on step 2.2 until your chatbot reaches 85% accuracy – in other words until it can understand 85% of sentences expressed by your users with a high level of confidence.

d) Build a concierge bot


Now that you’ve built the first version of your horizontal coverage, it’s time to put it to the test. This is where we introduce the concierge bot, which is a test bot into which testers enter questions, and that details what it has understood.
Testers can then confirm that the bot has understood a question correctly or mark the reply as false. This provides a second level of verification of the quality of your horizontal coverage.


Test your chatbot




e) Train again


The results of the concierge bot are then used to refine your horizontal coverage. Use the previously collected logs to enrich your intents until you again reach 85% accuracy as in step 3.

f) Roll your chatbot with horizontal coverage in production


As we’ve established above, horizontal coverage is just the first step in creating an effective bot. What makes a chatbot powerful is what it does after it has effectively understood the user request. And this means it’s important to automate business processes with the chatbot.

The horizontal coverage exercise helps you identify the topics that your users trigger most often, giving you data-driven insights into the topics you should automate first. If 50% of user questions revolve around creating a leave request, automating this process to simplify the user experience would seem to be a valuable business decision.

We recommend rolling in production with an initial version of the bot in which one-third of the requested use cases are automated and then gradually adding more automated use cases.
In next to no time, most common scenarios will be automated – with no interruption to your business.

How horizontal coverage positively impacts customer experience


We have seen great results with both internal and external customers after implementing the horizontal coverage in the bot building process.

Leading Swiss health insurance provider Groupe Mutuel wanted to make it easier and more convenient for clients to do business with them – all year round. The company implemented a chatbot that enables immediate responses to customers’ product-related questions in French and German via the Groupe Mutuel website – even outside office hours.
“We have been positively surprised by the fact that our customers are more than ready interact with the chatbot that we built using SAP Conversational AI and that 75% of their questions can be answered right away.”

David Cavalera, IT Project Manager, Groupe Mutuel SA

Internally, SAP managed a steady increase in queries with a chatbot built with SAP Conversational AI that provides real-time resolutions and a positive customer experience by answering technical questions across 10 lines of businesses from a variety of SAP resources.
“Built on SAP Conversational AI, Edmin is our intelligent expert within SAP S/4HANA Cloud User Community. Edmin provides capabilities that enhance the customer experience and reduce development efforts and costs within our support organization.”
Tony Johnson, Customer Experience Improvement, SAP SE

The SAP Conversational AI team is also collaborating with different lines of businesses within SAP to ensure horizontal coverage is integrated into the bot building process. Large teams of developers and project managers can create powerful enterprise chatbots from start to finish using the platform’s train, build, connect, and monitor capabilities.
Follow the easy steps presented in this tutorial to implement powerful horizontal coverage for your chatbot, enabling you to manage a large majority of your business use cases autonomously.

Do not hesitate to contact our team if you need to implement a chatbot in your business.
Feel free to ask your questions about any aspect of project methodology on SAP Answers or use the comment section below, our team will be there to help.

Part 2/3: How to design a perfect use case for your chatbot

Part 3/3: Launching your SAP Conversational AI chatbot effectively with Alpha, Beta, and Ramp up phases

TechEd session: Best Practices and Tips to Create Your Chatbot with SAP Conversational AI
2 Comments