What drove the decision to develop an AI-based use case?
At the beginning of 2023, we received several customer requests for the development of chatbots to answer frequently asked questions. We initially thought this would be easy to manage with existing solutions. But there were two main challenges:

• Either solutions quickly reached their limits. The answer "Sorry, I didn't understand that" has already caused frustration for many and causes potential customers to turn their backs on the company.
• On the other hand, modern large language models (LLMs) such as ChatGPT offer flexible answers, but sometimes provide wildly inaccurate information.

This motivated us to develop our own digital assistant: BennoAI BennoAI can be pre-programmed with answers to frequently asked questions. By combining an LLM trained with company data and the connection to existing systems, it is also able to provide correct answers to new questions.

What were your expectations or requirements for the AI application?
In addition to the accuracy and flexibility of the answers, data protection and transparency of the application were crucial for us. A good chatbot must be able to handle sensitive customer and company data. We guarantee that all data is processed in Germany and can even provide Benno on premise if required. Our tracking and analysis components ensure transparency in data processing and enable the data-driven development of further individual features of the digital assistant.

What frameworks/methods did you rely on for development?

We are a Silver Partner of IBM and use components from IBM Watson, which are also trusted by industry leaders such as SAP. These make our software solution secure and scalable, while at the same time allowing flexible integrations and extensions. The heart of BennoAI is a Large Language Model (LLM) trained with company data. There is no "one fits it all" solution for these models. Depending on the industry and use case, a different model (open source or not) may be a good fit. We maintain an overview of the models currently available and, thanks to our collaborations, can test and customise a wide variety of LLMs for companies.

The use of AI solutions must also be accompanied by a change in competences. How

have you dealt with this challenge in respect of your employees?
As software developers, we are used to the fact that working methods can quickly become outdated and that technology is developing at a rapid pace. Developments in AI have really taken off in the last year.

We were very pleased about this, as we have been working on AI-supported products for years and are now encountering much broader interest from customers. We are responding to constant change with dynamic and modular software structures that can be adapted quickly. This allows us to easily update or replace individual components and technologies as required.

This was also particularly important to us when developing BennoAI. We regularly work together with a wide variety of companies in the context of consulting. These consultings are the best motivation for us to stay up to date across industries, in addition to our regular internal briefings, further training and research. We are actively involved in organisations such as the Karlsruhe AI Hub and Bitkom in order to stay up to date and participate in shaping developments.

In what timeframe did they implement the AI solution in your company/your client‘s company?
The idea for Benno was born in April/May 2023 and we tested the first prototype in June 2023. Since then, we have been constantly developing Benno further, for example by adding new interfaces and integrations such as voice functionality. We are currently training Benno especially for internal use and for handling large volumes of documents, for example for law firms.

A conclusion to your AI solution:
With BennoAI, we have created a digital assistant that is customised, flexible and secure. Each "Benno" is precisely customised to the respective company and fits into the existing IT infrastructure. This sets our solution apart from existing SaaS products. A digital assistant that improves communication with customers and relieves employees of repetitive tasks.

One final piece of advice/tip to other entrepreneurs looking to apply AI:
We guide many organisations on their path to AI and have realised that an overview and getting people on board is the crucial factor at the beginning. This is why we often conduct consulting sessions and seminars in companies. Sometimes we find that the expectations of AI differ within the team. The first step is to create an understanding and consensus on what is possible. The next step can then be to implement a clearly defined use case and establish distinct key figures to measure success. From the very beginning, companies should focus on scalable solutions that can be expanded over time and integrated into the existing infrastructure.

https://sest.gmbh/

 

Picture: Sest GmbH