Private AI assistant for your documents
A private AI assistant is a conversational search engine connected to your own documents, and to nothing else. It answers from your procedures, past quotes, contracts and manuals, citing the document each answer came from. It does not answer from the internet, and your documents are never used to train a model.
What the subscription costs you
Six platforms in our catalogue sit in this box. The two closest to this need, Glean and Coveo, publish no price at all: the tier you would need is negotiated.
- Priced per seat per month, when the value of an assistant is precisely that everyone can ask it.
- Data-use terms differ by vendor, and change without consulting you.
- Your document library is uploaded to the vendor, into their index, under their control.
- Two of the six vendors in this category publish no pricing: the real number arrives after the sales call.
What we build
- An index of your documents, hosted where you decide, including in Canada on request.
- Sourced answers: every statement points back to the document and page it came from, so it can be checked.
- Your permissions enforced: an employee never receives an answer drawn from a document they cannot open.
- No model training on your data, written into the contract rather than a changeable policy page.
- Access with no per-user fee, because an assistant only three people can query is worth nothing.
The demo
A clickable screen, built with invented data, to show the shape of the system rather than describe it.
Private AI assistant
$10,000 to $28,000 · 2 to 5 weeks
Price and timeline
One-time fee. Hosting, updates, patches, backups and support from $129 per month, cancellable.
Work out what you pay today
Build your subscription stack and compare it with the cost of a system you own.
Rarely alone
Built alone, this system solves a problem. Built with the ones below, it stops passing the buck.
Project and job management
Procedures, plans and past mandates are exactly the corpus worth searching.
Helpdesk and ticketing
Most support questions have already been answered once, in a ticket nobody can find.
Client and vendor portal
Contracts and delivered documents, searchable, with permissions respected.
Frequently asked questions
Are our documents used to train a model?
No. Your documents are indexed to be searched, never used to train or fine-tune a model, neither by us nor by the model provider. That is a contract clause, not a line in an editable policy. If your sector requires that nothing leaves the country, the index and the processing can stay on Canadian soil.
Does the assistant respect our permissions?
Yes, and this is the part that has to be built properly. Permissions are applied at search time, not filtered afterwards: an employee cannot obtain, through a cleverly worded question, the contents of a document they are not allowed to open. The rules mirror the ones you already have.
What happens when the assistant does not know?
It says so. An assistant built on your documents must answer “that is not in the documents I have” rather than invent something plausible. That is the most important difference between this and a public chatbot, and it is what makes it usable for a procedure or a contract clause.
How many documents does it take to be worth it?
In practice, from the moment someone regularly spends ten minutes looking for information that already exists somewhere. A few hundred documents is enough. The deciding factor is not volume but how often people search and what a wrong answer costs.