A few words about Smartbroker: Smartbroker was launched in 2019 and is the first broker in Europe to enable customers to trade at special conditions. Customers can sometimes trade free of charge or place an order on the stock exchanges for very low prices. With its diverse range of products, Smartbroker offers the opportunity to invest across the board. In other words, Smartbroker actually offers the services of both a traditional house bank and an online bank — just without branches.
1. The tension in which Smartbroker was located: Maintaining high quality customer service despite very favorable conditions
With Smartbroker, wallstreet:online capital AG was in a position for the first time to provide organic growth with extreme scaling effects.
The challenge: to maintain both quantity and quality (in customer service) while offering very favourable conditions. Because: Smartbroker allows customers to make their investment decisions independently and process them more cheaply than with traditional house banks and online banks — but without having to make sacrifices in customer service.
Offering this is, of course, an area of tension. Not only because of the requirements just mentioned, but also because of the high competitive pressure in this sector.
The specific area of tension is between Smartbroker as a product provider, the expectations of customers (24/7 availability and immediate answer to the question) and the contest. It is therefore important to ensure that, despite the low price segment, the customer receives a full service that is satisfactory and does not cause doubt. In other words, the customer should not think that the low price reduces the quality of their service.
2. Guarantee scalability even as inventory grows
Strong customer growth means that the number of customer inquiries (especially at peak times) increases significantly. As a response to this, continuing to hire service personnel on a permanent basis is considered difficult for a number of reasons. On the one hand, due to Corona: “In the current situation, it is very difficult to find new employees because there are simply certain uncertainties on the market. Someone who currently has a secure job is not necessarily ready to change,” Thomas Soltau points out. And even if you find new employees, it would be difficult to offer them good onboarding — because the corona situation also makes it difficult to train new employees.
On the other hand, it also does not make sense from an economic point of view to continuously hire new service personnel:
“The more we grow, the more we scale, the more I have to pay attention to the workload in the workforce and keep an eye on the associated costs” — Thomas Soltau.
One of Smartbroker's goals is therefore to meet inventory growth by reducing the number of service personnel, at the same time reducing customer contact via telephone & email, but still offering an improved user experience and accessibility.
That almost calls for automation using an AI chatbot!
3. The two steps to an AI chatbot
1. Routing tickets & traffic
Requests can only be automated if they are actually made in the chatbot. Therefore, the elementary question starts with: How can inquiries be routed to the chatbot in the best possible way? Tip from Frederik Schröder:
“The company website is usually the best channel. There are usually existing and potential customers there who can ask their questions accordingly in the chatbot.”
In addition to the website, messaging applications such as WhatsApp or Facebook Messenger are also exciting options. Alternatively, a chatbot can be integrated into the company's own app. And referencing the chatbot via email or iVR can also be worthwhile. Ultimately, it always makes sense to welcome users with a chatbot where they are mostly active.
2. Automation in a chatbot via AI
moinAI's chatbot solution automates requests using various AI principles, the most important three being crowd learning, reassuring and dreaming.
Crowd Learning
Keyword: swarm intelligence. So that moinAI knows as many phrases and expressions as possible so as not to stumble across spelling mistakes, synonyms, etc., the AI is initially trained with a huge variety of variants. In concrete terms, this means that the AI is fed with thousands of sentences that come from real people.
Reassuring
that intelligent interrogation system from moinAI enables a user request to be answered correctly, even if the AI is actually uncertain. Because it often happens that users have a simple request, but formulate it in a very complex way. moinAI usually still understands what is meant, but also assures itself by asking, for example: “Do you mean that you want to reset your password?” If the feedback is positive here, it goes back as a strengthening signal and the subject area grows to include another possible formulation. This gradually creates ever better language comprehension.
Dreaming
If similar topics or phrases arise again and again, the chatbot automatically tries to recognize what is behind them, even if the topics and questions are not yet known. So if a topic appears more frequently, the chatbot notices that there is any relevance here and suggests the topics in question independently — this AI feature is called Dreaming. And this feature in particular is highlighted by Mohamed Dekhil, because it actually saves companies that use moinAI a lot of time. So you don't have to set up brainstorms, meetings, etc., to find new topics for the chatbot, as the chatbot simply suggests new topics itself. And not just any “out-of-thin air” topics, but topics that have actually already been asked about by users — i.e. are actually relevant.
4. A learning from the implementation: The chatbot was surprisingly well received from day 1
Since Thomas Soltau himself made this point, let's just let him speak for himself this statement from the webinar:
“Of course, we implemented the chatbot on the assumption that we would relieve customer service. But we didn't expect that it was accepted immediately after going live, almost from day 1 on, and was quite busy — that it would take away so much right away. That was surprising and it's easy to see that in the end, there are subtleties that go into this chatbot, that customers are also satisfied with them. On the one hand, because of the intelligent queries, which means that you simply have fewer negative associations that customers have with this chatbot. But for example, such a small subtlety as that when a customer asks a question, the answer doesn't come instantly, but the three points first. So the chatbot pretends to be written. (...) These are simply all topics that strongly ensure that customers accept it and make use of it very actively.”
5. Chatbot implementation: Not as complex and lengthy as you might imagine
During the webinar, a participant asked how long it would have taken to implement it at Smartbroker. The answer from project manager Mohamed Dekhil: “It actually worked very quickly, it took three weeks.” In other words, it took three weeks from contract conclusion to go-live. Shortly before going live, all you had to do was integrate a simple script, i.e. no relevant IT effort, says Mohamed Dekhil.
In the three weeks in question, two Smartbroker employees spent a great deal of time preparing the implementation, e.g. formulated the chatbot texts, selected the avatar, etc. However, if you were to put this in relation, with a view to how quickly the chatbot would have worked and how strongly the chatbot was accepted, you would very quickly have added value that could be derived from it, says Thomas Soltau. And further:
“That is truly remarkable. I wouldn't have expected that by far, neither that it would be so fast nor that it would take on so much load.”
That's why we'll just say a specific figure: Smartbroker's chatbot automates around 70% of all incoming chatbot requests.
Interested in more information and insights about AI chatbots?
Both for the use case customer service as well as for the use case marketing moinAI offers on-demand webinars, which use practical examples to show how an AI chatbot can be successfully implemented, how the AI principles work and what you should pay attention to in the “Chatbot Project.”
