Input
Call recording, transcript, support chat, sales notes, document set, knowledge-base file, CRM lead, or deal payload.
A company uploads a call, transcript, document, or CRM lead. The system runs RAG over the knowledge base, analyzes the conversation, scores the lead, extracts tasks, drafts a follow-up, asks for Telegram approval, and then creates a CRM task or comment through an adapter contract.
Call recording, transcript, support chat, sales notes, document set, knowledge-base file, CRM lead, or deal payload.
Ingestion, chunking, embeddings, retrieval, citations, summary, objections, risks, missing info, next step, and lead score.
Telegram approval flow for approve, reject, or edit draft before any CRM action is created.
Bitrix/CRM fake or sandbox adapter first, with a real contract for idempotency, retries, dead-letter handling, and audit.
State, RAG data, audit trail, retries, integration contracts, quality checks, approval history, and testable API behavior live in backend.
Docker Compose, CI, tests, screenshots, two-minute demo, runbook, metrics, and cost notes.
{
"summary": "what happened",
"objections": ["price", "timing"],
"next_step": "send follow-up with pricing options",
"lead_score": 82,
"risks": ["budget unclear"],
"missing_info": ["decision maker", "deadline"],
"tasks": ["create CRM follow-up task"],
"draft_follow_up": "message for human approval"
}
I use AI/Codex as an engineering accelerator, but the responsibility stays with me: architecture, backend state, tests, logs, deployment, runbooks, and quality of the shipped result.
The key distinction: n8n orchestrates events; the backend owns state, RAG, audit, retries, and integration contracts.
Backend/platform foundation: tenants, RBAC, audit/outbox, workflow rules, adapters, OpenAPI, Docker, CI, docs, and demo.
One-command offer demo: transcript, RAG retrieval, lead scoring, approval queue, n8n Telegram payload, and mock Bitrix CRM handoff after approval.
Fast review route for roles around AI automation, backend/platform, LLM workflows, CRM integrations, DevOps, and internal tools.