AI in Workflows

Orchard Core Workflows let you automate work without code. The AI features add two tasks that ask an AI model for an answer, and four events that start a workflow when something happens in an AI chat. Use them when you want a summary, a classification or a draft written automatically, or when you want your team to act on what a visitor told the website assistant.

Menu	Design > Workflows
Permission	Manage workflows
Feature	Workflows, plus AI Services. AI Completion using Direct Config also needs AI Chat Interactions.

Can't find it?

What you see depends on your role. If the menu, a button or a setting on this page is missing, ask your administrator to give you the permission listed above or to turn on the feature. Only administrators can turn features on and change site settings.

Add an AI task to a workflow

1. Open Design > Workflows and open your workflow, or click Create Workflow to start a new one.
2. Click Add Task, find AI Completion using Profile (or AI Completion using Direct Config) under Artificial Intelligence, and click Add.
3. Fill in the fields described below and click Save.
4. On the canvas, connect the task to the step before it, and connect its outcomes to the steps that follow.
5. Click Save.

The answer is stored under the name you give in Result property name, so later steps can use it. For example, with the name AI-Summary, a later task can insert the answer with {{ Workflow.Output["AI-Summary"].Content }}.

Every AI task has three outcomes:

Outcome	When it happens
Done	The AI answered. The answer is stored in the result property.
Drew Blank	The AI returned an empty answer.
Failed	The profile was not found, the prompt was empty, or the AI service returned an error.

tip

Connect Failed and Drew Blank to something useful, such as a notification to an administrator, so a workflow never stops silently.

AI Completion using Profile

Uses an existing AI profile, with its model, instructions, knowledge and tools. Use it when you already have a profile that does the job, for example a summarizer.

Field	What it does
Profiles	The AI profile that answers. Required.
Result property name	The name the answer is stored under. Start it with AI- so it does not clash with other tasks. Required.
Prompt template	The message sent to the AI. You can mix your own text with values from the workflow, written in Liquid, for example the output of an earlier task. Required.

AI Completion using Direct Config

Sets up the model and the instructions right inside the task, without a profile. Its editor has the same settings as a chat interaction, spread over tabs:

Tab	What you set
Content	Result property name, Prompt template, Orchestrator, Chat deployment, Utility deployment, System instructions, reusable prompt templates (when AI Prompt Templates is on), and the tuning values Max response tokens, Temperature, Top P, Frequency penalty and Presence penalty. Leave a value on Default to use the site default.
Knowledge	A data source to answer from, and documents to upload as the task's own knowledge (when those features are on). See Knowledge.
Capabilities	Tools, tool instances, MCP connections and agent connections the AI may use (when those features are on). See Tools and agents.

Start a workflow from an AI chat

Four events, under AI Chat, start a workflow when something happens in a chat. Each has one field, Profile: leave it on Any profile, or pick one profile to react only to its chats. Each event has one outcome, Done.

Event	When it starts
AI Chat Session Closed	A chat ends, because the AI detected a goodbye or because the visitor stopped replying for the profile's inactivity time.
AI Chat Session Field Extracted	The AI collected a value for one of the profile's data extraction fields, such as an email address.
AI Chat Session All Fields Extracted	Every data extraction field of the profile has a value. It fires once per chat.
AI Chat Session Post-Processed	The AI finished analyzing a closed chat with the profile's post-session tasks, such as a summary or a disposition.

The events pass the chat and its profile to the workflow, so later tasks can use what was collected. The exact values each event passes are listed in the Technical Manual.

Examples

• Send new leads to sales. A website assistant collects a visitor's name, email and question as data extraction fields. Start a workflow with AI Chat Session All Fields Extracted for that profile, and add an email task that sends the values to your sales team.
• Summarize closed chats. Give a support profile a post-session task that writes a short summary. Start a workflow with AI Chat Session Post-Processed and save or send the summary.
• Draft a reply to a form. When a contact form is submitted, add AI Completion using Profile with a profile that drafts friendly replies, and send the draft to a staff member for review.
• Tag incoming requests. Use AI Completion using Direct Config with instructions such as "Answer with one word: Sales, Support or Billing", then route the request on the answer.

For contact center events and tasks, see Contact Center Workflows.