Start building your own chatbot now!

Dear sponsor,

Innovation is pushing your business forward. “Innovate or Die” is something that Blackberry and Blockbuster didn’t get ten-fifteen years ago. So they died. Since then, we have all learnt the lesson. So did you. So you took the right decision and have decided to start a corporate chatbot project.

" A chatbot is NOT a “fancy feature”. It is here to push your business forward by generating revenue or reducing cost. Your project team can, and ought to, get you there."
— Omer Biran

I’m writing these lines to urge you to push your corporate  chatbot project team to bring you value for money. It is NOT a “fancy feature”. It is here to push your business forward by generating revenue or reducing cost. Your project team can, and ought to, get you there.

Disclaimer : The probability that you would find that I’m stating the obvious is not zero. I know. Personally, this kind of blogs drive me crazy. Nevertheless, in the last two years I’ve seen many CDOs budgeting and launching corporate chatbot projects without guiding the team to define sharp business objectives. The result? Either a chatbot that tries to do too much, “Mr. know-it-all”, has no identity whatsoever, where the user doesn’t understand a thing, gets frustrated, and the outcome is zero traction; or, a chatbot that does a great job, but not the right one, therefore generating low ROI and killing the project.

Step 1 : Fix the business objectives

Every chatbot project must start by fixing the business objectives of the chatbot. How would the chatbot generate a rapid ROI? Start with a high level objective and go progressively into the details.

For instance, a killer leverage point is productivity. Imagine a telecom operator operational context. The business objective definition could go down the following path :

  1. we want to reduce the costs of the customer support chat center
  2. the chat center cost depends on the number of incoming chats handled by agents
  3. by automating low value tasks, the number of chats handled by agents will decrease
  4. ticket topology analysis shows that 40% of incoming chats are about 11 different topics. (Pareto knew what he was talking about!)
  5. 5 of these top topics are:
    1. a user wishes to know when he’ll get his new box,
    2. report a network problem,
    3. inactivate a package option
    4. lost password
    5. pay invoice.
  6. The other 6 top topics are too hard to automate by the means of a chatbot. “I don’t understand my invoice” is not a good use case for a chatbot. A telecom operator CIO once told me that he himself doesn’t know how to explain their invoice, how would a chatbot know 🙂
  7. These 5 topics cover 20% of the overall incoming chats.
  8. Let’s be conservative and suppose that the rock star corporate chatbot would satisfy 50% of the users who engage it.
  9. This translates into making 10% savings on chat center budget

   We have a business case!


Step 2 : Set measures and KPI’s

As in any business project, to ensure the capacity to hit the business target and generate ROI, measures are to be put in place. The project team should define KPIs in order to measure the chatbot’s success.

A classic pitfall is to measure the conversation performance instead of the business performance. In the example used above, it is best to measure the number of inbound chats handled by human agents. That’s the key number that will eventually reduce costs.

I’m not saying that measuring the conversation performance is not important. Knowing whether the chatbot understands users well or not, whether the user is satisfied at the end of the conversation and whether his request was answered is clearly very relevant to our exercise. It’s just that these measures are not the bottom line. The bottom line is whether or not we hit the business objective. Other measures are simply indicators to guide use towards that goal! Start with the outcome in mind, and put in place your business objective KPI as the major dashboard.


Step 3: Choose the corporate chatbot use-case

The use case should be derived from the business objective, and not the other way round. So many corporate chatbot projects start with someone having a “dream” or a “feeling” or some “strong intuition” about the “winner use-case” or “the coolest chatbot ever”. It might indeed become an awesome chatbot with a sweet user experience with tons of likes and positive Facebook comments, but what does this chatbot do for your business?

Choosing the use case is a serious challenge, and is definitely not an easy task. The use-case, also called the chatbot’s identity, is one of the three factors that will determine the success of your chatbot among your users. (The other two being the user experience, and the NLP performance. If you’re using SAP Conversational AI, you’ll be just fine 😉 ). There are many great blog posts out there explaining how to design a kick-ass corporate chatbot identity, and you might want to take a look at these:


Step 4 : The C of DMAIC

Six sigma fans usually refer to 5 majors steps to optimize business processes, often shortened as DMAIC:

  • Define,
  • Measure,
  • Analyze,
  • Improve,
  • Control

Putting a corporate chatbot in place is not a one-shot exercise. The business objective KPI should be regularly monitored throughout both the Build phase and the Run phase of the project. The project team should verify constantly that the KPI converges to the objective as the project goes, and that once in target, it stays there! If this doesn’t happen, analyze and take action.


Breaktime : customer service and post-sales case study

Chatbots bring a great business opportunity to introduce cost efficiencies in the Customer Service world. This is probably not coming to you as breaking news. Although it is interesting to understand why this statement is true. It is basically driven by two forces: Threat and Feasibility.

Let’s start with the threat. Businesses encounter in this era of digitalization a huge peak of incoming customer demands. In particular via mail and chat. It is probably because we consumers become more and more exigent and hard to please. Combine that with the fact that complaining via mail or chat requires zero effort and you get the perfect recipe for 10x inbound customer support tickets. Why is it a threat? Because businesses cannot hire 10x customer support agents to maintain the quality of their customer service.

Now for the good news. It turns out that a great deal of customer care and post-sales tickets, industry cross, require a very short and simple conversation in order to provide the customer with a solution.

  • User: “My SIM card is blocked“
  • Agent: “What’s your PUK number?”
  • User: “75648392”
  • Agent: “Your SIM card is now unblocked”

It gets even better. Many of the complex customer issues involve a data collection phase in the beginning of the conversation (e.g., get the customer’s name and address): sounds like something a chatbot can easily handle! That’s feasibility 🙂

Conclusion in (almost) a single sentence

Chatbots are your new employees, they are the way to get your business successful, be demanding and set business objectives to fulfill. Do that and they’ll exceed your expectations.


PS: Thanks to Scott Adams for the comics, always on point. Check his work

Ask your questions on SAP Answers or get started with SAP Conversational AI!

Follow us on