Closed AI contour: one chat instead of ten systems
RAG and access inside your perimeter: roles, documents, CRM/ERP. Data stays with you. Discovery first — then a pilot on 1–2 sources.
What we deliver
One working chat
Staff ask what to do and where — answers cite your playbooks, orders and statuses, not generic model fluff.
Contour under your control
Server or VM, model/gateway, indexes, roles and access audit. On-prem or a dedicated Russian-hosted contour.
Tied to your systems
1–2 sources for the pilot: knowledge base, CRM, ERP, tickets. Expansion after a measurable KPI — not “connect everything at once”.
Your outcome
AskNayda
Hybrid search and vector layer over millions of cards — infrastructure for grounded answers.
PC Configurator
Dialog agent with orchestration and session state — production LLM sessions under control.
2–4 weeks
Typical pilot horizon: one chat, 1–2 sources, roles and a done criterion before expansion.
Technologies
Why NineLab?
Not a SaaS box: data and models stay with you
Same stack as cabinets and integrations — one owner through handoff
Pilot with KPI and eval, not a “watch ChatGPT answer” demo
1 chat
People lose time switching between cabinets, wiki and ERP. A closed contour pays off when routine questions close in one window without data leaving the perimeter.
Ready to discuss your project?
First consultation is free — describe your task and we'll propose architecture and timeline estimates.
How we work
- 1.
15–30 min: staff pain, data sources, security and contour (on-prem / dedicated).
- 2.
Discovery 1–2 weeks: pilot scope, roles, risks, timeline and budget range.
- 3.
Pilot 2–4 weeks: index, chat, 3–5 roles, eval and audit before prod.
- 4.
Launch, quality and cost monitoring; next sources as a separate stage.
What you get
- Working chat and RAG pipeline in your contour — code in your Git.
- Diagram: sources, roles, access, human escalation points.
- Eval set and answer-quality metrics before scope expands.
- Docs for security and the process owner.
Pricing and timeline
Discovery — from 80,000 ₽ (1–2 weeks). Closed-contour pilot on 1–2 sources — separate quote after Discovery, usually 2–4 weeks. Retainer for indexes and quality — by agreement.
We do not promise a “chatbot for 50k”. On-prem, critical infrastructure and full ERP sync are scoped separately.
Need a chat in your contour — not a demo on a public API
Briefly describe systems and security requirements — we return a Discovery and pilot range.
Discuss your project
Describe your task — we respond within 2 business hours.
Free initial consultation
Timeline and architecture assessment
NDA on request
Or call: +7 (912) 232-74-87 · full contacts page
Related reading
Closed AI contour: FAQ
Data and indexes stay in your perimeter, with roles and access audit. Answers ground in your sources — not a public chat that leaks context.
No. A cabinet is the order and accounting process. The AI contour is a staff layer on top of systems. They often follow the same Discovery.
Not always. A dedicated contour and an approved-model gateway can be enough. Hardware and model are fixed in Discovery.
Discovery — from 80,000 ₽. Pilot and infrastructure — separate quote after sources and security review. We do not quote a hard pilot number before access.
Usually 1–2: playbooks/wiki and one operational contour (CRM, tickets or an ERP slice). “All of ERP at once” is a later stage.
After Discovery, a one-chat pilot on 1–2 sources is usually 2–4 weeks.
