Artificial Intelligence From Bots – A Profile of e-bot7

Artificial Intelligence From Bots – A Profile of e-bot7

The Munich startup e-bot7 develops and integrates artificial intelligence into companies’ existing customer service systems.  The system analyzes incoming messages, forwards them to the right department and gives support agents smart answer recommendations while they are working. This reduces the processing time by up to 80% and recurring questions are automated. We spoke with Xaver Lehmann, one of the founders.

e-bot7 combines two trending topics — bots and artificial intelligence (AI). How do you assess the market potential in this area?

The market potential for artificial intelligence solutions in the field of customer service is enormous. The majority of current CRM systems are completely outdated. Customer service staff spends most of their time answering repetitive questions instead of concentrating on more demanding issues. This inefficiency can be alleviated with help from AI and chatbots.

The market potential for virtual chatbot assistants is enormous, too. Forrester forecasts a thousand-fold increase in their market volume by 2020. Our first target markets include the telecommunications industry, insurance companies and banks. That makes for a market volume of 25 billion euros.

Can you give us a brief overview of the different bot solutions?

There are different technical approaches to constructing a bot. Most bots that you encounter on Facebook Messenger or in chat windows on websites are called “rule-based bots.” They are usually built using existing frameworks and are manually fed with data, aka questions & answers. This data creates what is known as a static database, because the entered questions & answers do not change. These bots work well as long as the customer asks a question that is already in the databank.

If a question is asked, however, that is not in the databank, the customer receives an error message. At the end of each day, the service staff has to go through the system to analyze the error messages and manually enter the right answers. First, it is impossible for a static databank to cover every possible question a customer might ask. Second, no company wants to invest more resources to train the bot at the end of the day.

The three e-bot7 founders: Fabian Beringer, Xaver Lehmann and Maximilian Gerer (from left to right)

So what do you do better than other bots?

At e-bot7, these are precisely the problems we solve. Using artificial intelligence and neuronal networks makes it possible to understand the meaning of incoming questions. To make sure our artificial intelligence does not forward incorrect answers to customers, we offer a hybrid agent and AI solution.

Hybrid agent and AI solution

What does that mean exactly?

The company first selects the necessary level of certainty, for example 98%. This allows the company to determine how certain their AI needs to be before sending an automated answer. As soon as a customer sends a question, the AI system supplies a recommended answer.

If the AI system is more than 98% certain that the recommended answer matches the question, it then automatically sends the answer to the customer. If the AI system is only 95% certain, then it sends the recommended answer to an agent, who then checks the recommended answer.

The agent then has three options. First, the agent can send the recommended answer, which trains the AI system. This means that if the same question is asked again, the answer will have a higher degree of certainty. Second, the agent can modify the recommendation and then send it. This also trains the system by adding the modified answer to the knowledge base. Third, if a completely new question has been asked, the agent can create, save and send the new answer. This process creates a databank that constantly adapts and learns from the operational processes performed by the agent. This allows the AI system to accumulate new knowledge from every interaction with the agent. At the same time, the AI system is monitored by the agent through supervised learning.

“The results even surprised us”

Your AI system seems to learn very quickly. Could you share some figures with us?

When dealing with AI and machine learning, it is important to remember that the system needs data and time to learn quickly and effectively. AI learns faster from 10,000 questions each day compared to 100. The results from our technology even surprised us.

In one of our case studies with a large telecommunications company, our bot was already sending more answers (61%) than the agent after just two months. Focusing on the German language has made it possible for us to quickly train other European languages, namely within four weeks.

What are your three secrets to success?

  • Our vision is really big. Within the next three years, we want to become Europe’s market leader in AI and machine learning applications. That objective motivates us each and every day to work even harder and with more focus.
  • A strong team that you can trust and share success with. It is possible to successfully found and grow a company together.
  • Perseverance and courage: we don’t let critics or nay-sayers get us down. You find plenty of them in B2B business. We take criticism seriously and learn from it quickly.

What is a current tricky issue for you?

Our team is growing very fast. So it is a major challenge at the moment to find the right employees who fit into the team perfectly and are just as motivated as we are as the company founders.

Focus on the added value

What do you think is the decisive factor for sustainable development of a company?

For us, there is not just one decisive factor for the sustainable development of a company. However, one of our most important factors for success is the focus on our value proposition. We need to have a clear picture every day of how, when, where and for whom our software provides added value. Without keeping that in mind and believing in your own product, it is not possible to be successful over the long term.

Fail forward – what mistake has taught you the most?

During our first round of financing, we were too confident about everything going according to plan. Then at the very last moment, one of our investors bailed out for no reason. We fortunately had a very strong network of investors at that point in time who supported us. What we learned was that a deal is only final after everything has been signed (note from editor: learn more about the financing round here).

What advice do you share with other founders?

It is worth working for your dream 24/7.

“Munich is the ideal startup city”

And what advantages do you think Munich offers founders?

Munich is the ideal startup city. With the exception of the high rent prices, Munich has the best infrastructure in Germany for founding a company. First, you have very strong universities such as LMU and TUM with their entrepreneurship initiatives like MakerSpace or the LMU EC. Those are places where founders with complimentary skills can meet up and start to implement their ideas.

There are also so many VC’s and accelerators in Munich who are able to supply initial venture capital.  Last but not least, Munich is simply the most beautiful city in Germany.

We agree with Xaver, of course, and would like to thank him for the interesting interview and insights!

An article by

Munich Startup

Munich Startup ist das offizielle Startup Portal für München und die Region, das von der Landeshauptstadt München entwickelt wurde. Mitinitiatoren bzw. Kooperationspartner sind die UnternehmerTUM, das Entrepreneurship Center der LMU, das Strascheg Center for Entrepreneurship (SCE) und die IHK für München und Oberbayern. Träger ist die Münchner Gewerbehof- und Technologiezentrum GmbH (MGH).

By continuing to use the site, you agree to the use of cookies. more information

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.

Close