Case: RAG-based AI chatbot for a website
Hurja is a software company that delivers on its word. Our team of more than twenty passionate professionals develops AI-enhanced software, mobile apps, and web applications with uncompromising expertise and a solution-oriented mindset.
A content-rich website needs a smarter way to serve visitors
Many organizations have a wealth of valuable content on their websites, such as service descriptions, customer stories, blog posts, instructions, and answers to common customer questions. This content supports customers at different stages of the buying journey, helps them compare options, and builds trust in the company’s expertise.
Website visitors’ expectations have changed: people increasingly want to search for information in natural language by asking direct questions. A visitor may want to know, for example, whether a service fits their situation, whether the company has previous experience in their industry, or how the collaboration would work in practice. A RAG-based AI chatbot gives the website a way to serve visitors in the moment and helps them use the site’s content more naturally.
The same starting point applies to our own website. Over the years, Hurja’s website has accumulated a lot of content about our services, expertise, customer work, and ways of working. We wanted to improve the website service experience, make the site more interactive, and help visitors move from searching for information toward getting in touch.
A RAG-based AI chatbot helps website visitors use the site’s content
We built a RAG-based AI chatbot for our website so that visitors can ask about the topic they need in their own words and receive an answer based on the site’s own content. The chatbot helps users find information about our services, references, AI solutions, technologies, industry experience, and project process.
The solution makes the website more interactive and serves visitors at the moment their need for information arises. The website chatbot also acts as a practical example of the kind of AI solutions we build for our customers to support information discovery, customer service, sales, and smoother internal work.
How does a RAG-based AI chatbot work on a website?
A RAG-based AI chatbot retrieves answers from a selected knowledge base and forms a response based on the content it finds. In our website chatbot, the knowledge base consists of the published content on hurja.fi.
A website visitor can ask the AI chatbot a question in natural language. The question can relate to services, references, AI solutions, technologies, industry experience, or how a project typically progresses.
The user does not need to know the right search term, service name, or which page contains the information. They can approach the topic based on their own need and ask, for example, whether the company has experience in a certain industry, how AI could be used, or what kind of solution would fit their situation.
The RAG chatbot forms its answer based on the organization’s own content. First, the bot searches for relevant information from sources such as the website, instructions, documentation, or other materials. Then the AI chatbot forms a response based on the content it has found.
The solution uses the RAG method and a vector database in the background. RAG, or Retrieval-Augmented Generation, is a method in which a language model first retrieves information from selected content and then forms an answer based on it. The vector database helps find semantically relevant content even when the user’s question and the source text use different wording.
The AI chatbot can be guided to respond in the organization’s own tone of voice, so the answers support the brand’s way of communicating. This is especially important on websites, in customer service, and as part of sales support, where the bot should not only deliver information but also feel like a natural part of the company’s service experience.
In our case, the bot uses the content published on our website. In a customer solution, the knowledge base could include product documentation, user guides, onboarding material, internal process descriptions, or customer service guidelines.
The AI chatbot can show the source on which the answer is based. On a website, this can mean a service page, customer story, blog post, or another content page.
Showing sources increases transparency and supports the user’s journey on the website. The user gets a quick answer and can continue reading the original page if they want to explore the topic in more detail.
Not every question can be answered from the selected content. When the knowledge base does not contain enough information, the AI chatbot communicates this clearly and guides the user forward, for example toward contacting a person.
This is an essential part of a reliable chatbot solution. When the bot is used on a website, in customer service, or for internal information search, it must also recognize situations where a human or more detailed information is needed.
When new content is published on the website, it can be made available to the AI chatbot through indexing. This allows the bot to keep up with new customer stories, blog posts, service pages, and instructions.
This makes a RAG-based AI chatbot a flexible solution. The same structure can grow with the organization’s content and serve different use cases without rebuilding the entire solution.
Making better use of existing information with an AI chatbot
A RAG-based AI chatbot helps organizations make better use of existing content and improve the user experience in digital channels. On a website, this can mean that a potential customer finds relevant services, references, and additional information more easily before getting in touch. In internal use, it can help an employee find the right instruction without browsing through several systems or materials.
Key benefits include
- faster information discovery in natural language
- a more interactive and service-oriented website experience
- more effective use of existing content
- more reliable answers based on the organization’s own material
- stronger support for sales and customer service
- a scalable structure for different materials and use cases
Our website chatbot demonstrates in practice how a RAG-based AI chatbot works in a real environment and how the same approach can be applied to your company’s own needs. The solution is especially well suited for organizations that have a lot of useful content, but where finding, using, or sharing that information takes time.
A RAG chatbot project can start with a focused pilot where a clear use case, suitable material, and the goal of the solution are defined. You can currently try our test-stage solution on the right-hand side of the website. You can also give us feedback on the bot’s functionality, answers, and user experience. Because the solution is still in development, there may occasionally be room for improvement, for example in Finnish-language answers.
If your customers or employees repeatedly search for the same information, let’s explore what kind of AI chatbot would best serve your users and your business.
Shall we get started?
"*" indicates required fields
INNOCUM