Frequently asked questions — and what we do not gloss over.
The questions that come up before a decision, including the uncomfortable ones. Where an answer only becomes clear on your own documents, we say so — and how we check it before we start.
What it actually is
The four questions that are the best place to start.
What is an AI knowledge base — in one sentence?
A store for your company’s documents that you ask questions in plain language, and which names the source as a numbered reference for every statement in an answer. Clicking the reference opens the original document.
We already have ChatGPT or Copilot — what is the difference?
The knowledge base is tied to your own documents: a separate instance per company, permissions by role and document type, and for every statement the source in your documents. If it finds nothing there, it is designed to say so openly. The two are not mutually exclusive — for general writing tasks a general assistant can still make sense. Whether an existing tool is enough for your questions is best shown by comparing them on the same questions in a demo.
Does it replace our ERP or our industry software?
No, and it is not meant to. Your ERP records what was sold, ordered and invoiced — it stays the leading system for that. The knowledge base is for how things were done: instructions, service reports, solutions to problems, project knowledge, correspondence. Whether documents from your programs can be brought in, we check before we start; the plan is read-only access, nothing is changed there.
What does it explicitly not do?
It sends nothing on its own and does not automate workflows. On request it creates drafts — Word, Excel, PDF or an email draft (in the Premium package) — but you have to send them yourself. And what is nowhere in your documents, it cannot answer; then it says that it found nothing.
Documents, formats and systems
What goes in, how it gets in and what happens to your existing documents.
Which file formats are processed?
The common office formats: PDF, Word, Excel, PowerPoint, text and CSV files as well as images and scans (JPG, PNG, TIFF). Plus emails with attachments (.eml/.msg), ZIP archives and voice memos. Scans and photos go through text recognition (OCR), older originals too. How readable your documents are, you see in a demo; handwriting is the hardest case.
Do we have to tidy up first?
No, folders that have grown over the years are the normal case. The search is designed to look for meaning and not just the file name — the maintenance contract for system 3 should come up even if the file is called “Scan_0047.pdf”. What that looks like with your documents, a demo shows. The only useful thing to decide beforehand is which areas should go in at all.
Do we have to upload everything?
Not necessarily. Where a connection to your system is possible, it is set up and new documents come in through it — whether that works for you, we check before we start. For everything else, and until a connection is in place, there is the upload: files, whole folders, emails, photos and voice memos.
What about our old program?
A connection needs a way in which data may be read: an interface, a file store or a scheduled export. Whether one of these exists, is enabled and can be reached depends on version, vendor and network — we clarify that in advance, before any effort is incurred. Sometimes a scheduled export is the more robust route than a live connection. Nothing is changed in the program itself.
How are documents classified?
When documents are captured, the system suggests attributes: customer, project, installation, date, topic. The document type is confirmed by a person in the review area, because it decides who may see the document — we do not leave that to automation.
We have a large paper archive — will that hold up the start?
The archive does not have to be fully digitised before you start. The usual approach is to start with the digital documents; scanned folders are added bit by bit, the most important first.
Who sees what, and where the data is stored
Questions about server location, access rights and deletion.
Does everyone see everything?
No. Permissions depend on document types and roles, not on individual files: payroll stays with management, the maintenance manual is visible to the whole team. A cut-off date is possible too — only from a certain date, or only older documents up to then. If your role lacks the permission, the system says so instead of pretending there is nothing.
Where are our documents stored?
In a separate instance per company on our servers in Germany — no shared system, no mixing with other companies’ documents. The language model runs in a data centre in the EU. If files are on a server at your company, they are mirrored in encrypted form into your instance; the instance itself does not run at your premises.
Are our documents used for training?
No. Your documents are not used to train AI models; this is guaranteed by contract, and the data processing agreement (DPA) is included.
Does our IT have to open a port?
No. The plan is exclusively outbound, encrypted connections from your network to your instance. No open port, no remote access, no access to your server from outside. Whether and how this can be implemented at your company, we clarify with your IT before we start; there is a technical description for this.
Does the system go on the internet by itself?
Only if you allow it. Web search has three levels: not allowed, approval in the chat for the individual question, or allowed. At the lowest level, answers come exclusively from your own documents.
What happens when we delete something?
Deleted documents are removed from the search index and permanently deleted after ninety days — immediately on request.
What if it does not know something?
If the system finds nothing on the question in your documents, it is designed to say so openly — that is not a guarantee against errors. Where it answers, every statement carries a numbered reference that leads to the source — so every answer can be checked against the original.
Introduction and everyday use
What lies ahead for your company before the first question is answered.
How long does the introduction take?
From kickoff to acceptance, six weeks of support are planned. How long the setup takes before that depends on the volume of documents and the sources to be connected. You receive the concrete schedule with the offer.
What do we have to change in day-to-day business?
Your folders, your mailbox and your programs stay as they are; the knowledge base runs on our servers. Your part is a few appointments agreed in advance: kickoff, introduction, a meeting in week four and the acceptance.
Do we need our own IT skills or an IT department?
Not for running the knowledge base — we take care of set-up, updates and backups. Using it is similar to a search engine: type or dictate a question, read the answer, open the source via the reference. We show how it works in the introduction. If an IT service provider looks after you, we coordinate the connection with them.
Will our technicians really use it?
That cannot be promised in advance, but it can be checked. Two things make a big difference: whether the interface is handy on the phone and whether the first questions go to the real documents. That is why the introduction comes after the import, not before, and the interface can be added to the phone’s home screen. In the first six weeks we look at what is being asked and what is missing.
What happens if the answers are not right at first?
That is what the support is for: we look at real questions from your company and fine-tune. At the kickoff we record twenty questions together that the acceptance is based on; in week six we go through together what the system does with them.
Who keeps the system up to date later?
New documents, dictated notes and feedback are added in everyday use. There is a contact person at your company for this. If we notice that something is missing, we get in touch.
Price, contract and exit
What it costs, what that is based on and how you can get out again.
What does it cost?
A flat fee per company, from €500 per month, plus a one-off setup. The price depends on the sources, not on the number of users. All amounts are net. The packages are listed in detail on the pricing page.
Why not per user?
Because with a per-head licence usually only part of the team gets access. The technician on site who only wants to look something up now and then misses out — and calls the office instead. With a flat fee per company, that question does not arise.
How many employees can take part?
As many as you have. The number of users has no effect on the price. Who may see what, you control via roles and document types.
What does connecting our system cost?
One-click connections via your account — Microsoft 365 with SharePoint, OneDrive, Teams and Outlook, Google Workspace and Dropbox — are included in the setup; we check the prerequisites in advance. For other systems we check before we start whether there is a workable route; their connection is set up and billed by effort — what it costs is fixed beforehand. A database connection also comes with ongoing support: monitoring and adjustment when the program is updated.
Can we see it first?
Yes. In a thirty-minute video demo we show you examples from your industry: how questions are answered and what the sources look like. Before we start, we also go through the answers on real questions from your company.
Who owns the data?
You do, throughout. You remain the controller under data protection law; we are the processor. When the contract ends, you receive an export of your documents, after which they are deleted.
What applies if the acceptance in week six is not passed?
At the kickoff we agree on twenty questions together – along with the documents the answer should come from. At acceptance we go through them together and look at what the system does with them. What follows from that is set out in the offer.
Your question is not listed?
Then ask it in the first call — or ask the system itself: in a demo there are examples from your industry, and you can add your own documents. Technical details for your IT are under Security, the terms under Pricing.
You see more on your own documents than here.
Thirty minutes by video, with examples from your industry. No obligation.