Real-time customer support with scalable AI chatbot
About the client
8.2 Certification is a German B2B company that assists its customers in A, B, or C-type certifying products and services that generate renewable energy, including wind, photovoltaic, and biogas systems.
Challenge
For B2B service companies, high-quality and ongoing customer support is a critical growth driver. Traditional approaches such as static FAQ pages and human-only support teams are no longer effective in competitive markets. Rising support costs, slow response times, and limited availability directly impact customer experience and retention. Users today expect immediate, reliable, and continuous communication they can trust.
With this in mind, 8.2 Certification engaged us to develop an AI-powered FAQ chatbot designed to modernize and scale their customer support. The solution needed to address the following challenges:
Ensure continuing customer support without workforce strain
As customer inquiries grew in both volume and complexity, 8.2 Certification found it difficult to maintain reliable 24/7 support solely relying on the static FAQ section. This prompted the company to seek a scalable solution capable of providing continuous assistance without significantly expanding its support team.
Enhance user experience
Customers now expect clear and personalized responses, without having to navigate complex website structures or ask repeated questions. To support this experience, the client needed an AI chatbot with a well-structured, continuously enriched knowledge base and a user-friendly interface.
Ensure seamless chatbot integration with the existing website
The chatbot had to operate smoothly within the client’s existing CMS-based website without disrupting performance or user workflows. It was crucial to enable consistent communication between the chatbot and website content while maintaining a unified user experience.
Improve customer consultation management
Managing follow-up consultations, feedback, and customer insights manually was time-consuming. The client needed a centralized admin panel to review and manage customer requests, analyze search behavior, and continuously refine chatbot responses.
Delivered value
24/7 customer support through automation
The chatbot now automates 70–80% of routine customer inquiries, providing instant responses at any time. This ensures uninterrupted customer support and scalability of business operations as demand grows.
Reduced support team workload and increased productivity
By taking over repetitive FAQ-related questions, the solution reduced employees’ workload by 30–50%. Support teams can now focus on higher-value consultations, improving both productivity and service quality.
Improved customer experience
The AI chatbot handles the common questions, while complex or unanswered questions are escalated to the management team. A centralized admin panel enables teams to review and handle consultation requests, capture customer feedback, and analyze search behavior. This ensures no inquiry is lost and allows continuous refinement of chatbot responses based on real user needs.
Solution
The project started with the client extending their existing Q&A knowledge base and sharing it with our team. To validate the concept early, we developed a proof of concept (PoC) and a demo application, including a basic front end, to test the FAQ chatbot behavior on real data and confirm the optimal technical approach.
Following PoC approval, the next step was to design the chatbot UI/UX and implement the data science layer using a retrieval-augmented generation (RAG) architecture. Our engineers uploaded the knowledge base dataset into OpenAI to convert this data into multidimensional vectors and make it usable for semantic search — the chatbot's core functionality that enables users to receive immediate responses.
To further train the model and create more datasets with potential questions and answers, we integrated synthetic data generation and collected users’ feedback. Our data science engineers also foresaw cases in which the database didn't have the right answers to customer questions. For this, the chatbot has built-in thresholds that give users hints, like rephrasing the question or contacting the manager if no correct answer is available at the established confidence level.
Once the data science pipeline was finalized, we implemented the backend logic and integrated the chatbot into the client’s CMS-based website. The system communicates with the frontend via REST APIs, ensuring stable performance and seamless user interaction. The final phase introduced a custom admin panel that allows internal teams to manage chatbot content, review consultation requests, analyze search history, and continuously update the knowledge base.
For future implementations, the chatbot can be integrated with more hints to make it clear to users what types of questions they can expect responses to.
Technologies
jQuery
FastAPI
PostgreSQL
Kubernetes
OpenAI SDK
React
Wagtail
REST API
We are thrilled with our project outcome and had a great experience working with Lemberg Solutions experts throughout the project. They delivered everything on time and as expected, and we are glad to continue cooperating on the following projects.
Volodymyr Lysak
CEO at 8.2 Certification GmbH
How it works
Initiation of conversation
The chatbot greets the customer and asks how it can assist.
Customer request submission
The customer enters a question or request through the chat interface.
Automated response delivery
The chatbot delivers an immediate, context-aware answer based on the knowledge base.
Response validation and continuation
If the answer is sufficient, the customer may continue the conversation by asking follow-up questions.
Escalation to human support
If the chatbot cannot provide a confident or satisfactory response, the customer can leave contact details to request a personalized consultation with a manager.
Feedback and continuous improvement
Customers can rate their experience after each interaction. Unresolved or uncommon questions are recorded, allowing administrators to expand the knowledge base and improve future responses.
