Workflows & Automations
The Q-AI Bot is a set of named background automation workflows, each with one job. Together they carry a chat from "AI turned on" to "handed back to a human" and keep each project's knowledge in sync. This page is the roster: what each workflow does and where its behavior is documented in full. Refer back when a teammate mentions one by name.
How the workflows fit together
At a high level there are two clocks ticking:
- The conversation clock: driven by chat events. When a chat turns the AI on, sends a message, or gets handed back to a human, a workflow reacts in real time.
- The knowledge clock: driven by schedules and file changes. In the background, other workflows keep each project's knowledge base in sync so the AI always answers from current information.
Conversation workflows
These are the part of the system a support hire will think about most.
AI Assistant - Main
The core reply loop: the workflow that does the actual answering, decides what happens next (continue, limit, or hand back), and keeps each conversation's context across turns.
See How It Works for a turn-by-turn walkthrough and Conversation Lifecycle for how a chat starts, continues, and ends.
AI Assistant - Dev Sandbox
The safe twin of AI Assistant - Main: the same job, used to trial changes before they go live so new behavior is validated without touching real conversations. It also writes each test turn into an evaluation report, making it a natural place to check answer quality.
See Reporting & Evaluation for what gets logged during these test runs.
AI Abandoned Bot
Handles idle conversations: walks a per-project re-engagement ladder of timed steps when someone stops replying.
See Abandoned Bot System for the full ladder mechanics.
Knowledge workflows
These are what keep the answers accurate: each one feeds a different content source into the knowledge base the AI reads from.
AI File Management
Handles the files a project provides. When a knowledge file is added, changed, or removed in the project's document folder, this workflow mirrors that change into the AI's searchable index. It is also the path that updates a project's system prompt when the prompt document changes.
See Knowledge Files for how files flow into the knowledge base.
Scraping Websites for AI - Main Loop
Handles a project's website content, working from a curated list of page URLs (not by crawling) and routing each page to one of three small helpers.
See Website Scraping for the full mechanism and the per-page helpers.
Setup & insight workflows
These run around the edges of a conversation rather than inside it.
AI New Customer Onboarding
Provisions a brand-new AI project end to end (the empty knowledge base, configuration, document folders, default prompt and evaluation report, and the lifecycle scenarios) behind the Onboard AI Bot tool.
See Onboarding a New Project for what gets provisioned.
Update AI Report Sheets
A small logging helper that AI Assistant - Main calls on every turn, in the background. It appends the turn's question, the AI's answer, and the model's reasoning to the project's evaluation report (for a human to score later), and records usage for cost and volume tracking. It is "fire-and-forget": the conversation never waits on it.
See Reporting & Evaluation for how the logs are structured and read.
Where to go next
- Conversation Lifecycle: how a chat moves through the AI and back to humans, plus the Texter scenarios that wake these workflows up.
- Per-Project Settings: the configuration each of these workflows reads.
- Scenario Marketplace: import the Q-AI scenarios (search
q-aior filter theai-bottag) that trigger the conversation workflows.