What changes for your team
Every company has questions that always land with the same few people: Which travel costs are reimbursed? Which delivery terms apply to this customer? Where is the manual for this machine? With a chatbot built on your documents, those questions go there instead, from the first working day after launch.
The answer comes in seconds, with a link to the passage it comes from. The person asking can check it. The colleague who used to answer keeps working.
What an AI chatbot is, and what we do
An AI chatbot is a program your team asks questions in plain language, the way they would ask a colleague. Ours answers from your company's documents, cites the source for every answer, and says so when your documents have no answer. It replaces digging through folders and asking the same few people again and again.
We handle everything it takes: we load your documents, set up the access rights, check the answers against your real questions, and extend the chatbot one team at a time.
Where the answers come from
The chatbot answers from the documents you give it: policies, manuals and product documentation, as PDF, Word or Excel files. Before every answer it finds the matching passages and answers only from those. The method is called RAG; our guide What Is RAG? shows how the search works.
RAG is not training. The model learns nothing from your documents; it reads the matching passages again for every question. That is why a new or updated document is available as soon as it is added.
What we need from you
We start with one team and its most frequent questions. Other areas follow once that part works. To start, we need:
- the documents that team works with
- a list of real questions, each with the correct answer
- someone who checks answers during the test phase
- the rule for who may see which documents
How we check the answers
Your list of real questions is the benchmark. Before the chatbot goes live, it answers every one, and we check two things: did it find the right passage, and does its answer match that passage. Every later change is measured against the same list again.
When your documents have no answer, the chatbot says so instead of guessing. Those questions are collected. After launch they show you which knowledge is missing and which page to write next.
Everyone sees only what they are allowed to see
The chatbot follows the access rights that apply in your company. Someone who cannot open a document gets no answer from it either.
No model is trained on your data. If it must not leave your network at all, the language model runs on your own servers too; our local AI page shows what that looks like.
What it looks like: Aven
Aven, our own product, works on the same principle. It answers from your company knowledge, marks the source down to the line, and works in German, Turkish and English, hosted in Germany and the EU. The demo on our home page shows a question being answered.
You start with a pilot for one team, with a fixed scope and a cost you know before you commit. If Aven does not fit your case, we build the chatbot for you.
Chatbot, AI agent or AI integration?
A chatbot delivers answers. If it should also act on them, for example create a ticket or prepare an order in your system, it becomes an AI agent. If software you already use should get an AI feature of its own, that is AI integration.
Frequently asked questions
How much does an AI chatbot cost?
It depends on whether Aven covers your case or a chatbot is built for you, and on how many sources are connected. With Aven, model usage is included in the monthly price. The Project Wizard gives you a first range, and you know the cost of the pilot before you commit.
What does the chatbot do when it finds no answer?
It says your documents contain nothing on the question, and it does not guess. The question is recorded so your team can write the missing page.
What happens when a document changes?
The new version is added, and from then on the chatbot answers from it. Nothing is retrained.
Where are the documents the chatbot answers from stored?
In data centers in Germany or the EU, or on your own servers. Your data is not used to train AI models. For data that must not leave your network, the whole system runs on your premises.
Can the chatbot answer our customers' questions too?
Yes. It answers customers at any hour, only from documents released for customers, and hands anything it cannot answer to your team. If it should also create tickets in your CRM, it becomes an AI agent.