AI platform
Every model for your staff — without leaking company data
The NAVXNET AI platform gives staff the leading models in one window and gives the company control: who uses what, how much they spend and what data leaves the company.
Corporate access to AI models
Staff use the leading models in one window — no personal subscriptions, no VPNs, no work files sent to third-party services. The administrator decides who can use which models and sets limits and a monthly budget.
- GPT
- Claude
- Gemini
- local models
Administration
- Access to models and agents by role and department
- Per-employee limits and a total monthly budget
- Alerts at 80% and 100% of the budget, with an optional hard stop
- Audit log of administrator actions
Leak protection
Before text goes to an external model, the platform finds personal data in it and replaces it with tags. The model answers using the tags, and the real values are put back on your side.
“Draft a reply to Alina Karimova, tel. +992 90 123 45 67, about order 5521”
“Draft a reply to [NAME_1], tel. [PHONE_1], about order 5521”
A finished reply with “Alina Karimova” and her phone number back in place
- Names, phone numbers, emails and other personal data are replaced with tags before anything goes to an external model, and restored in the answer on your side
- A company dictionary: names, clients and terms that never leave the company
- Sensitive tasks run on a local model, so data stays on your servers
- Writing to 1C, email and other systems only after a person confirms; emails stay as drafts
- Internal addresses for integrations only from the administrator’s allow-list
- Two-factor sign-in is mandatory for administrators
Agents that know your instructions and work inside your systems
A platform for corporate AI agents. Each agent gets its own knowledge base, access to the systems it needs and its own permissions. Staff use it in a chat or through a form; software uses it through an API.
- Instructions
- What to do, what tone to use and what not to do.
- Knowledge base
- Company documents: product manuals, contracts, policies.
- Integrations
- 1C, email, databases, spreadsheets, external services — only those the administrator has allowed.
- Interface
- A chat for staff, a form for routine tasks or an API for software.
Already in use
- Answers to marketplace customers
- Questions and reviews on OZON and Wildberries: the agent finds the answer in product manuals by model code and shows the source.
- Replies to claims
- A legal agent drafts the reply using the company template and produces a DOCX and PDF document with an outgoing number.
- Staff assistant
- Chat over internal documents, web search, recognition of scans and voice messages.
Knowledge base
Search across company documents by meaning and by keywords, with re-ranking. Answers link to the source; scans and voice messages are recognised.
In product catalogues, search takes the model code into account, so answers don’t mix up similar models.
Integrations
- 1C
- Databases
- Google Sheets
- HTTP API
- Web search
An open API lets you build an agent into any software.
Local models on your servers
For tasks where data must not leave the company, models run on your GPU servers. We select the hardware and set everything up turnkey.
- Long documents in full: contracts, correspondence, reports
- Recognition of scans and voice messages runs locally too
- One platform: staff choose the model, the protection rules stay the same
- Role-based access to agents, two-factor sign-in, spending limits and an audit log of administrator actions.
Let’s discuss your task
Tell us what you want to monitor or automate. We’ll reply, show you the product and prepare an estimate.