Many products, many customer inquiries: MicroNova uses innovative AI technologies to make support processes more efficient. An AI-supported chatbot makes it possible to answer product questions in the shortest possible time - efficiently, precisely and comprehensibly.
MicroNova uses such a solution successfully in its own company: As the exclusive sales partner for around 80 ManageEngine products in Germany, we receive technical questions about the various IT management solutions on a daily basis. While the sales team can answer general questions, such as licensing, directly, specific questions often have to be forwarded to the support and consulting team - a time-consuming process. To optimize this workflow, MicroNova now relies internally on an AI-supported chatbot.
The basis of chatbot responses: structured information
In order for AI to provide well-founded answers, the right data must first be made available. In our use case, we use our website manageengine.de as one of the central data sources for the AI: MicroNova provides comprehensive information on all ManageEngine products on around 1,000 subpages on this website. Internal documentation, databases, support tickets or product manuals can also be used as data sources.
To make the website data usable for the AI, we developed a script that reads all subpages and extracts the text content.
The content is divided into smaller sections and stored in a so-called vector database. Each section is also provided with meta tags that indicate, for example, which product the section belongs to or whether it is a feature description.
This structured approach makes it possible to search specifically for relevant information without having to search the entire database. This significantly improves the efficiency of the query.
Data protection & security when using AI chatbots
ChatGPT from Open AI serves as the technical basis. ChatGPT is integrated via Microsoft Azure. In order to meet German data protection requirements, the data is processed exclusively in the Swedish region. This ensures that the solution is GDPR-compliant and adheres to the highest security standards
How the AI chatbot works
The chatbot works according to a clear principle: every user question is transmitted to the AI with a suitable context. This ensures that the answer is based exclusively on the information provided.
As an alternative to the context-based transmission of information, the fine-tuning of an AI model can also be considered. However, it is not advisable to have the answer output on the ChatGPT database: The level of knowledge may be outdated and there is a risk of inaccurate or incorrect answers.
With the context-based approach, this process follows when a user enters a question in the chat interface:
- Comparison with the vector database: The AI searches for sections that match the content of the question. Synonyms are also recognized, for example "report" and "report".
- Filtering by product and keywords: If the question contains a specific product such as "ADAudit" or a specific topic such as "editions", the search in the database is narrowed down accordingly.
- Limiting the amount of data: To optimize the response quality, the AI only receives a limited number of relevant sections as context.
- Answer generation with source citation: The AI formulates a precise answer and supplements it with a source citation to ensure traceability for the user.
Chatbot for customer support: practical example
For our use case, we ask the question "Does ADAudit support preconfigured reports?" and use a prompt to instruct the AI to answer the question based on the attached context. We also give the instruction to output the answer "I don't know" if the question cannot be answered based on the context. This avoids so-called hallucination, i.e. the invention of answers by the AI.
The five text passages taken into account are listed at the end of this answer. This allows responsible team members to verify the results of the AI.
The advantages of the AI chatbot at a glance
An AI-supported chatbot can significantly improve the provision of information in companies:
- Faster response times: Sales can answer simple technical questions independently without having to call on support.
- More efficient use of resources: Support teams are relieved and can concentrate on more complex issues.
- Higher response quality: AI provides relevant and precise answers thanks to targeted data queries.
- Traceability: Each answer contains a source reference so that information can be verified quickly.
Versatile: chatbot, self-service solution, knowledge management and much more
The chatbot can not only be used internally, but can also be made available to customers as a self-service solution. Users can receive answers to frequently asked questions directly via the website - straightforward and without waiting times.
In addition, the solution can be easily transferred to other use cases, for example as a reference work for internal company guidelines on travel expense reports, company car use, etc. The answers are available in a matter of seconds, thus relieving the burden on internal contact persons.