> For the complete documentation index, see [llms.txt](https://oodlz-ai-studio.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://oodlz-ai-studio.gitbook.io/docs/welcome/how-it-works.md).

# How It Works

The high-level flow — Inbound on every channel, Outbound on a monthly cadence, Reporting once a month, all surfaced in the workspace.

> **Producers run autonomously, approval gates preserve editorial control, the workspace is the daily-use surface.**

The system has 3 production cycles operating on different cadences. Each runs autonomously and writes to the workspace where the brand reviews, approves, or just reads.

## The Inbound Cycle

Every inbound message — comment, DM, chat widget, voice — hits one webhook. The agent classifies intent before any expensive call fires. Active product intent gets the full commercial response. Casual compliments get a 1–2 sentence reply with no tool call. Harmful input is dropped silently. Real questions get answered from the brand's knowledge base. Real escalation requests trigger a full handoff to support with conversation context attached.

The cycle runs every time a message arrives. No batching. Cost and latency are kept low by the cheap classifier and posture routing — bad-faith input never reaches the main agent.

## The Outbound Cycle

Once a month, the Topics producer discovers and scores \~200 candidate topics for the brand. The brand approves the ones to turn into articles. Per approved topic, the Articles producer runs a research-write-format-publish pipeline and produces a draft website article in minutes. The brand approves the draft. The Repurpose producer schedules the article for the next available Monday/Wednesday/Friday at 9am, generates 3 brand-native images, atomises the article into 5 platform-native social posts, and dispatches everything via Zernio.

2 approval gates sit in the cycle — Topics before article generation, Articles before publish. Between gates, the producers run autonomously.

## The Reporting Cycle

The Reporting Agents run on the 2nd of every month at 06:00 UTC. Each connected channel — SEO via GSC, web analytics via GA4, social via the platform APIs — has a dedicated workflow that pulls the prior month's data, calculates the month-over-month delta, generates a 3-bullet AI summary, and writes everything to the workspace. The dashboard renders within an hour of the cron firing.

## The Architectural Backbone

Every brand runs on dedicated infrastructure: own n8n instance, own workspace, own vector index, own image rendering account. No shared workflows, no shared data, no shared state across brands. See [Multi-Tenant Isolation](/docs/architecture-and-stack/multi-tenant-isolation.md) for the full architectural picture.

![Architecture overview — three production cycles (Inbound, Outbound, Reporting) all writing to the workspace where the brand reviews, approves, and reads](https://2935174940-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F9WhvO5Y38RJO9oNIdz9G%2Fuploads%2Fgit-blob-650b2ed55d3199630083ee25c345e8a68051f689%2Fdiagram-architecture-overview.svg?alt=media)