Workflow assistants
Workflow assistants are an advanced form of AI assistants. Instead of a fully open conversation, the user goes through a defined series of steps driven by AI. This makes it possible to guide complex, multi-step processes — from start to finish, step by step.
What is a workflow assistant?
A workflow assistant works based on a visual flowchart of nodes connected by arrows. Each node performs a specific task:
| Node type | Description |
|---|---|
| Start / End | Starting point and endpoint of the workflow |
| Evaluate | Analyzes the context and stores information in the workflow memory |
| Condition | Evaluates an expression and routes the flow in the right direction |
| LLM | Calls a language model for a specific task (optional: waits for input) |
| Approval | Asks a yes/no question and routes the flow based on the answer |
| Pass through | Connects routes without an AI task or visible output |
| Subagent | Performs the same task in parallel for multiple items |
The workflow automatically saves its state after each step. When the user provides input, the workflow resumes exactly where it left off — even after the browser is closed.
How do I use a workflow assistant?
Start a new conversation and choose a workflow assistant from the list. The assistant automatically guides you through all steps. In some steps, the assistant waits for your response; in others, it continues working independently without input.
On the right side of the chat input, you will see the Workflow button. Click it to open a panel with all steps and their current status (waiting, active, done, waiting for input, skipped, error).
You can close the chat at any time and continue later. The workflow will resume from where you left off.
Example: In-depth research
The preset In-depth research is an example of a workflow assistant available in all environments. It guides the user through a thorough research process:
- Define the topic — the AI asks for the research topic and desired depth
- Formulate search questions — the AI creates focused search questions
- Conduct research — the AI performs the research and collects information
- Generate report — the AI writes a structured research report
- Wrap up — the user reviews the report and can ask additional questions
Creating a workflow assistant
Open Assistants and choose the Workflow assistants tab. In the card at the top, click New workflow and choose Create or edit workflow with AI, Blank workflow, n8n workflow or a preset. Under Quick start, you can start a preset directly or choose Use as template.
You can also convert your own assistant: in the assistant card menu, choose Open as workflow. You will get a new, unpublished draft; the original assistant remains unchanged. Review and test the suggested steps before publishing.
Workflow assistants, like regular assistants, can be shared with user groups and colleagues, also from the editor using the Share button in the toolbar. When shared with colleagues, AI-Public can automatically send an email notification so recipients know the workflow assistant is available.
Starting an n8n workflow
In addition to AI-driven workflow assistants, AI-Public can also start an n8n workflow via a webhook. An n8n workflow does not open a chat: without fields, the webhook is called directly; with fields, a form opens first. Administrators create this via Assistants → Workflow assistants → New workflow → n8n workflow.
Read the full explanation on the n8n workflows page.
General settings
When creating or editing a workflow, you configure the following for the entire workflow in the settings panel:
- Model: The AI model used by all steps.
- Tools: Which tools are available (e.g. Recording, Internet search). Steps can disable globally enabled tools per step.
- MCP servers: External integrations.
- Documents and images: Files that LLM steps can consult.
Creating a triage node
An evaluate node analyzes the user's input and stores structured values in memory (e.g. route, topic). Configure:
- A system prompt describing what needs to be analyzed.
- The output schema: fields with name and type (
string,number,boolean,enum).
The stored values are available as state.fieldname in condition nodes.
Setting up a condition node
A condition node reads the workflow memory and routes the flow to the correct next step using CEL expressions:
| Desired behavior | Expression |
|---|---|
Field is true | state.sufficient_info == true |
| Field has a value | state.route == "report" |
| Numeric threshold | state.score > 7 |
Each branch has its own connection to the next node.
Using a pass-through node
A pass-through node does not perform an AI task itself. The node simply passes the workflow on to the next step. Use this node to keep routes organized, for example when a path skips a step and then rejoins the main route.
A pass-through node must have at least one incoming and one outgoing connection. The node is not shown as a separate content step in the workflow progress.
Recording in workflows
Enable the Recording tool in the general settings and add an instruction in a step’s system prompt to record results. The user can view the recorded documents via the Recording button (📄) in the toolbar.
Restoring a conversation
Not happy with the direction of the conversation? Using the restore button on an earlier message, you can roll the workflow back to that point. All messages after that are deleted and the workflow resumes from that exact moment.
Restoring is permanent — deleted messages cannot be recovered.
Files and subagents in workflows
A workflow can ask the user to provide files, for example a source document, evaluation model, contract, policy document, or completed form. The user can upload a new file or choose an existing file from File Management. Files added through a workflow form become available to the assistant in that chat, but are not automatically selected for regular chat questions.
New files are processed first before the workflow can continue reliably. If a required file is still being processed or has failed, the user must wait or choose/upload the file again.
The Subagent node performs the same task in parallel for multiple items. Use this when a workflow needs to process a list, for example one document, participant, employee, customer, case file, or form per run. The subagent receives the current item each time in an isolated context. Therefore, make prompts explicitly refer to the current item and avoid instructions that process multiple items at once.
Creating or editing a workflow with AI
Discuss your workflow with AI. Valid changes appear in the editor. Review and test the result before saving or publishing.