One person knows everything and gets interrupted all day
We turn their knowledge and documents into one searchable base. AI answers first and hands difficult cases to a human.
AI that answers from your documents and shows its source
We build a knowledge base from your policies, contracts, manuals and price lists. Staff and customers ask in plain language and get an answer with a link to the exact document.
Quote and schedule after your brief
Who it's for
- Companies where new hires spend months learning where things are documented
- Support and sales teams that answer the same questions every day
- Legal, finance and technical teams sitting on large file archives
- Businesses that want a website chat answering from their own documentation
Problems we solve
Files are scattered across Google Drive, Notion, email and PDFs, and nobody knows the latest version
We connect the sources, remove duplicates and mark current versions. Every answer links to the file and page it came from.
We worry the AI will invent an answer and mislead a customer
We restrict it to answer only from the base. If the documents do not contain the answer, it says so and passes the question to a person. We test it on your real questions before launch.
We do not want internal files going somewhere unknown
We show where data is stored and which model processes it, and we set access rules for who can ask about what. All accounts and files are yours.
What's included
- Document audit: what goes into the base and what stays out
- File preparation: OCR for scans, duplicate cleanup, version labels
- Splitting texts into chunks and indexing them in a vector database (search by meaning, not just keywords)
- Hybrid search: by meaning and by exact terms such as contract numbers or SKUs
- A language model set up to answer only from retrieved passages
- Source citations in every answer
- Interface: website widget, Telegram bot, Slack or Microsoft Teams
- Role-based access: different document sets for different users
- A test set of 30–50 real questions and a quality report
- A short guide for the person who will keep the base updated
Choose the scope
Use a package to describe what you need. We agree on the quote and schedule after reviewing your brief and requirements.
01
Pilot
Testing the idea on one area, such as support or HR
- Up to 100 documents from one source
- One channel: Telegram bot or web chat
- Test on 30 questions against an agreed KPI
- Answers with source links
02
Working base
Teams that need a base for a whole department or a website
- Several sources: drive, Notion, email, PDFs
- OCR for scans and hybrid search
- Two channels, for example website and Telegram
- Roles and access rules
- Handoff to a human when AI cannot answer
03
Company-wide
Large archives, several departments, strict security needs
- CRM or ticketing integration
- Automatic re-indexing when files change
- Department-level access control
- Hosting on your server or your cloud account of choice
- Dashboard showing top questions and gaps in documentation
How we work
Brief and requirements
You describe the task and current workflow. We clarify requirements, choose a first stage and prepare a quote.
Pilot KPI
We agree on a measurable target up front, for example how many of 30 typical questions AI must answer correctly with a source.
Building the base
We clean files, run OCR, index the texts and configure search and the model.
Testing on your questions
You and your staff ask real questions. We see where the AI fails and fix the search settings or the documents themselves.
Launch and handover
We connect the channel and train the person who will maintain the base. All access and files go to you. Payment is in stages.
Scope and limits
- We do not promise error-free answers. Quality depends on your documents, so we measure it on a test set of questions.
- We cannot create knowledge that is not in your files. If rules live only in people's heads, they need to be written down first.
- We do not build a base that keeps your documents tidy by itself. Someone on your side has to update files and remove outdated ones.
FAQ
What is RAG in plain words?
RAG (retrieval-augmented generation) works in two steps. First, the system finds the relevant passages in your base. Then a language model writes an answer from them and cites the source. So the AI relies on your files, not on general knowledge.
How is this better than uploading files to ChatGPT?
For a handful of files, that is often enough. With hundreds of documents you need cleanup, search by meaning and by exact terms, access rules and quality testing. You also need your own channel: a website, Telegram or an internal chat.
What do we need to provide?
Access to the documents, 30–50 real questions from staff or customers, and a person who can confirm which answers are correct. You also decide which files must stay out of the base. We will tell you exactly what to prepare.
Is our data safe?
Together we choose where the base is stored and which model processes it. The system can run on your own server or cloud account. Access rules ensure people see only the documents they are allowed to see. For EU clients we account for GDPR requirements.
What does it cost to run after launch?
We prepare a quote after reviewing your brief. We list support and any third-party costs separately: subscriptions, APIs, hosting or media spend where the project needs them. Provider rates are checked before work starts.
Can we start small?
Yes, the pilot starts We take one area and one channel and measure the KPI on real questions. If the result works, we expand. If it does not, you found out for a small budget, and you keep all files and access.
Start with your task
Describe your business, the outcome you need and your current workflow. We will clarify the requirements and prepare a quote after your brief.