Close-up: One language in two weeks, but no ego or: Why is AI not explained in the ‘Sesame Street’?
4th December 2019 – Thomas Wilke from 42Cap interviewed Fabian Beringer, CEO e-bot7, about the internationalisation plans for our AI-based bot model, the development of AI in Europe, the war for German talent, fear of progress and risks, and his biggest mistakes.
We got to know e-bot7 CEO Fabian Beringer and his team in the course of our second fund. After their latest round of funding, I talked to him about the internationalization plans for the AI-based bot model, the development of AI in Europe, the war for German talent, fear of progress and risks, and his biggest mistakes. By the way, fittingly with Fabian’s insights, we shared our view on e-bot7: in our Why We Invested In series.
Thomas Wilke: You’ve just received 5.5 million in funding in your Series A round. Tell us what you plan to do with the money. And what was important to you for that round?
Fabian Beringer: It was important to us that we didn’t need the money to survive. That was the case in the Seed Round as well as in the recent Series A. Currently we have raised the capital to not only be a player in Germany and DACH, but to establish ourselves internationally. When raising funds, we also try to take the perspective of an investor, just as we do with our customers, asking ourselves what they want and what they see in us. It is important for us to know what somebody has in mind with our company. Do they have to sell their shares after three to five years, for example, in order to achieve their fund goals? With e-bot7, we don’t want to be just a number in a giant pool, we want to have partners who burn for our product and are there when we need them. As for the round before we also wanted investors who understand our situation. Often it is not the exact number in the report that counts, just as the share price cannot be meticulously predicted. In spite of all the measuring, it has to fit on a personal level.
Your internationalization is interesting, because your product can be used in countless industries, but the bot has to learn the language of the market. How quickly can you do this, and what range of conversation situations do your chatbots cover?
Our algorithms today able to learn a language within two weeks and even conduct more complex dialogues. We currently offer all European languages as well as Japanese. Right now we are focusing on expansion in Europe, and the US market will follow soon. Our AI technology covers everything from standard questions such as “How much does it cost?” to more complex questions such as complaints and damage claims. The product is therefore relevant for almost every customer contact, from betting providers to the automotive and travel industries to banks, insurers and financial service providers. It is live with Deutsche Bahn, Miele, Commerzbank and o2 — from external customer service to internal help desks. What is crucial is to know the individual pain points of a company or segment.
Your start after university was pragmatic and gave you a great advantage for product development. Strictly speaking, these were the workshops you held for companies.
Yes, Xaver and I worked on a bot application case in addition to our university exams and didn’t initially think of a start-up but rather of an agency business. We were exceptions at the University of Maastricht doing that already. Because the goal of 95 percent of the people there was to start at McKinsey & Co, alternatively in investment banking or in a highly remunerated corporate job. There were questions like ‘Do you see yourself more in supply chain management or consulting?’ Of course, these are exciting topics, but we wanted to take over and design more than just one area. As we got closer to graduation, Xaver and I sat together over a beer and said ‘Let’s start something’. Two days later on the train we read a tech article in which Facebook announced ‘We’re making bots now’. So we looked for a useful application case and started to build e-bot7 parallel to our exams. With our chatbot-workshops on Eventbrite we could cover our running costs, but most valuable to us was the insight into the problems and use cases in customer service departments of organizations. When it was clear to us which direction our product should go, we brought in Max, our co-founder and CTO. He built the first version of the product in a very short time. No one else could have done it like that.
“The user shouldn’t have to deal with AI.”
What convinced Alex and me was how you built your Agent+AI model. It supports employees from day one, learns from them and is thus also supervised. Can you describe the reasoning behind this approach?
The process to find the right answer is different in every company, as is the data and preparation time a company has to train an AI. Our product can be trained on data sets beforehand, as well as used for a cold start. By suggesting answers to employees and being corrected by human colleagues, the system can learn without making mistakes. This is because the bot can only respond independently if, for example, it is 97 percent certain. The larger idea behind this is a plug & play approach. Our platform can be integrated into any company landscape within four weeks, and the team can customize it individually in no time at all. I don’t need a data science team or a technical developer for this at all. The user should not have to deal with AI, but rather be able to use the solution immediately. Since very few companies are prepared for the topic of artificial intelligence, we wanted to enable a start without preparation.
There are also companies that believe in ‘Build it yourself’. You’ve always relied on your plug-and-play approach for pitches. Can share your experience on this matter?
We’ve always had contact with companies who wanted to build it themselves. But almost everyone of them called us again after the project had more or less gone up against the wall because it wasn’t as easy as they thought or because they ran out of resources. Conversely, in the beginning we often heard ‘This is not going to happen, the AI is not going to develop as quickly’. Then you need someone who believes in you. Our first customer o2 was someone like that. From then on and ever since, we have remained very close to each of our customers. Many start-ups are very focused on themselves and their product vision. We find it very important to work with customers as partners and get a lot of feedback. In our opinion, this is the only way to build the best product on the market.
“We have a huge problem in Germany because we don’t have Open Data Pools.”
Could you sketch out how you can advance your vision and what drives you in doing so?
In the US, a big vision is mandatory to get money at all. In Europe you are likely to be throttled with big goals. You need to defend your vision against such pushbacks and simply go on. The point, however, is that you must evaluate the vision precisely and reflect enough to know whether it is still coherent. As far as AI is concerned Germany is stuck in a mode of hesitation, not only compared to the US, but also to China. Unfortunately, there is some void for AI development and companies here in terms of data availability and public funding. There is still some change needed here!
You co-founded the Bundesverband für KI (Federal Association for AI) in 2018. In your opinion, what are the biggest hurdles in AI development and what are your goals?
In Germany we don’t only have too few start-ups, but also far too little funding from politics and a huge problem because we don’t have open data pools that we need for research and development. For example, if I want to train an AI in image recognition of diseases, I need a critical amount of images. With our bot product, you need a critical mass of customer service tickets to train also account for deviations. The algorithms have to, for example, understand confusing conversations, spelling mistakes and interspersed foreign words such as English or Turkish expressions in German. We are talking about valuable structural data, anonymized, so without a name.
The Bundesverband KI was founded last year to drive AI in Germany. We now have about 220 members, including startups, data scientists and AI experts. When making our selection, we pay great attention to a concrete AI focus. Today, many products have an AI label on them that shouldn’t.
“Germany is the free training ground for companies like Google, Amazon and Facebook.”
Why don’t you tell us about your experiences with large companies and corporations, besides your customers? How do they react to the topic of AI?
Many German companies often get stuck in the fear ‘AI cuts jobs’. Most medium-sized companies do not yet have an applicable AI strategy. The crazy thing is: Germany is partly the free training ground for companies such as Google, Amazon and Facebook. We, for example, would like to hire more German developers at e-bot7. The problem is: Many students who get a scholarship, e.g. for Stanford, just stay in the US and go to Google and Co. The goal must be to make Germany and Europe the leading AI ecosystem.
German Gründerszene has written about you that you educate between ‘Terminator or Utopia’. How do you see the situation between real dangers and fear of progress?
AI has long been omnipresent in our everyday lives, whether Spotify suggests something to me or my phone searches out all the pictures for the keyword café. At the same time, the dangers are clear, for example through surveillance via face recognition as in China or fake and masked applications like the FaceApp. In Europe, we need to establish criteria for transparency and standards in programming, data documentation, all these things. We have a chance in Europe and Germany, for example, to establish an ‘AI Made in Germany’. What we want to advance are companies that use AI for positive scenarios and according to our democratic beliefs. But what I see again and again at conferences is that horror scenarios sell well. Speakers take advantage of this and make a brand out of it. What we need, however, is clarification and education on the subject of artificial intelligence. In other words, why isn’t AI explained in the Sesame Street?
“You have to eliminate your ego completely.”
As an entrepreneur, leadership and self-conception are among the biggest personal topics. How do you deal with it?
During the growth phase, it is important to transfer both the presence in the markets as well as the corporate culture from HQ to these markets. At the same time, you have to have the best product. Here, a strong team is the be-all and end-all. People who have the right skills or can quickly adapt and acquire them. At the same time, people need to have the right mindset and have fun driving things forward. It is important as a founder and team member to remain down to earth at all times. You have to completely eliminate your ego. It’s not about egos, it’s about substance. If you just want to make a name for yourself, show that you are better, you are doing something wrong. It has to be about the cause and must not become personal. We are happy about every piece of advice and reflect a lot, but we also put ourselves in a different perspective. We want to enjoy our vision. That is the balancing act between ‘I am confident’ and ‘I remain realistic’. But fun is a good key. With fun hybris is out of question.
Which mistakes have you personally learned the most from?
I used to be doing projects in the music business. We released tracks and at some point we had a concept for an artist, including video and branding. Then we had a pretty good lead. But the contract partner wanted to use a YouTuber instead of a new, unknown artist. I didn’t like the idea and the whole thing came to nothing, which I regretted very much. That doesn’t sound like a spectacular mistake at first, but for me it was one of the most important learnings: Don’t run in circles on details. Because it’s only by doing stuff that you come to the point where you can realize your own vision.