Skip to content
saitrix
RU

The route from brief to launch

Every project rides the same four stations. You always know where the train is, what arrives at the next stop, and what we need from you to keep moving.

  1. Map

    Days 1–3

    One structured call and a short questionnaire. We map how inquiries reach you today, where they stall, and which segment automation pays back fastest. Honest scoping happens here: if a smaller package covers your case, we say so.

    What you receive

    • Funnel map with the leak points marked
    • Recommended package and integration list
    • Delivery plan with dates

  2. Prototype

    Week 1

    Before any production code, you click through the actual conversations: how the assistant greets, qualifies, answers from your documents, and hands off. Wording, tone, and escalation rules are approved here, when changing them costs nothing.

    What you receive

    • Clickable conversation flows on your real scenarios
    • Knowledge base draft from your documents
    • Voice samples on your scripts (voice projects)

  3. Build & connect

    Weeks 2–3

    The approved prototype becomes a production system: knowledge base finalized, CRM wired, channels connected, telephony configured. Test conversations run against edge cases you supply and the awkward ones we bring from experience.

    What you receive

    • Working system on your live channels
    • CRM integration writing real records
    • Test protocol with your sign-off cases

  4. Run & tune

    Week 4 and after

    Launch is a beginning. For the first month we review real conversations weekly: what customers actually ask, where the assistant hesitates, which flows need a second exit. The system you have at day 30 is measurably sharper than at day 1.

    What you receive

    • Weekly tuning sessions for the first month
    • Conversation analytics dashboard
    • Documentation and handover materials

What the systems run on

Boring choices where reliability matters, current models where intelligence does.

  • Models

    Current-generation language models matched to the task and budget; open models for on-premise installs. We benchmark on your real questions, not vendor demos.

  • Retrieval

    Every assistant runs on retrieval-augmented generation over your indexed documents. Answers trace to sources; confidence gates the handoff to humans.

  • Rails

    amoCRM, Bitrix24, Telegram Bot API, SIP telephony, and REST integrations with your systems. Local rails first: the tools businesses here actually run.

  • Operations

    Logging from day one, conversation analytics, and durable storage of every inquiry. If a dialog went wrong, we can always show why.

About working together

How much of our time does delivery take?

Plan for three to four hours total across the project: the audit call, prototype review, and launch sign-off. Document collection is asynchronous; we work with what you have.

What do you need from us to start?

Access to the documents customers ask about (price lists, terms, catalog) and one person empowered to approve wording. CRM access comes later, at the build stage.

What if the prototype shows this will not work for us?

That is the prototype's job. You stop at the prototype stage having spent a fraction of the budget, with the flows and knowledge base draft yours to keep.

Who owns the system after launch?

You do. Knowledge base, flows, integrations, and documentation are handed over in full. The support plan is optional, not a hostage situation.