Velaro Workflows Guide
Velaro Workflows Guide
What is a workflow?
A workflow is an automated sequence of steps triggered by an event. Workflows run without an agent — they handle chat, collect information, route to teams, call APIs, and send responses automatically.
Use workflows to automate: lead capture, support triage, appointment booking, FAQ answers, data lookup, and CRM updates.
Creating a workflow
1. Go to Workflows in the left sidebar.
2. Click New Workflow.
3. Choose a trigger (e.g. New Conversation, Incoming Message, Schedule).
4. Add nodes by clicking the + button below any node.
5. Click Save, then Activate to make it live.
Workflow triggers
| Trigger | When it fires |
|---|---|
| New Conversation | When a visitor starts a new chat |
| Incoming Message | When a message matches a keyword or condition |
| Schedule | On a cron timer (e.g. daily at 9am) |
| Webhook | When an external system POSTs to the workflow URL |
| API Call | When called from another workflow or integration |
Node types
Send Message — sends text, image, or card to the visitor. Supports variables like <<firstName>>.
Ask Question — asks a question and captures the answer into a variable. Set the variable name under "Save answer as".
Set Data — stores a value into a variable. Use this after collecting data to save it to the contact record.
Condition — branches the workflow based on a value. Example: if <<accountTier>> equals "Enterprise", go to branch A; otherwise branch B.
Route to Team — sends the conversation to a human agent team. Workflow ends here and the chat enters the queue.
HTTP Request — calls an external API. Supports GET, POST, PUT, DELETE. Response fields can be saved as variables.
AI / Bot — sends conversation to an AI bot for a response. The bot uses your configured knowledge, system prompt, and Bot Profile (persona). You must have at least one Bot Profile created under Settings → Bots → Personas before this node will work correctly. See the section "Bot Profiles and the AI/Bot Node" below.
Send Email — sends an email via the configured email channel.
End — terminates the workflow. Use this at the end of each branch.
Using variables in workflows
Variables let you personalize messages and make decisions based on contact data.
<<firstName>>— contact's first name<<email>>— contact's email<<plan>>— any custom field you created (use the field name)
See the Variables Guide for full syntax and scope rules.
Testing a workflow
1. Open the workflow editor.
2. Click Test in the top toolbar.
3. A test chat window opens — interact with it to trace each node.
4. Check the Execution Log tab to see which branch ran and what values each variable held.
Activating and deactivating
- Active: workflow runs for all matching triggers.
- Inactive: workflow is saved but not running. Use this for drafts.
- Click the toggle at the top of the editor to change the state.
Workflow templates (Starter Wizard)
Each integration page (HubSpot, BigCommerce, Shopify, etc.) has a Launch Starter Wizard button. This creates a pre-built workflow for that integration in one click. Start there instead of building from scratch.
Common workflow patterns
Collect email then route to agent
Ask Question (email) → Set Data (save email to contact) → Route to Team
VIP fast-track routing
Condition (<<vipStatus>> equals true) → Route to VIP Team | Route to Standard Team
API lookup before responding
HTTP Request (look up account by email) → Set Data (save account tier) → Send Message (<<accountTier>> plan greeting)
FAQ auto-answer with fallback
AI Bot → Condition (visitor still needs help) → Route to Agent
Scripted workflows vs AI workflows
You can build a workflow with zero AI, or blend AI into any part of it.
Pure scripted workflow — use Send Message + Ask Question + Condition + Route to Team. Every response is authored by you. Good for: regulated industries, strict compliance scripts, simple lead forms, onboarding sequences.
AI-assisted workflow — add an AI / Bot node anywhere you want flexible, conversational handling. Good for: open-ended support questions, knowledge base lookups, natural language intent routing.
Hybrid — scripted nodes up front (collect structured data), then AI node for the conversation. Best of both worlds for most support use cases.
You do NOT need Dialogflow, LUIS, or any external NLU service. The workflow structure handles deterministic routing; the AI node handles natural language when you want it.
Bot Profiles and the AI/Bot Node
Before using an AI/Bot node in any workflow, you need a Bot Profile.
A Bot Profile is a reusable character layer — it defines who the bot is (name, avatar), how it speaks (tone, greeting), what voice it uses (ElevenLabs), and which LLM model it runs on. Create profiles at Settings → Bots → Personas.
How the AI/Bot node uses a profile:
1. Add an AI/Bot node to your workflow
2. Select a Bot Profile from the dropdown (or leave blank to use the Default profile)
3. The node inherits the profile's tone, greeting, voice, and AI model automatically
4. Your system prompt in the node adds role-specific instructions on top of the persona
Why use separate profiles?
- One "SalesBot" profile for sales workflows, one "SupportBot" for support — different tone, model, and voice for each
- Swap a model or voice across all workflows by editing one profile, not every node
- The Default profile is auto-used by nodes that don't specify one — always keep a Default configured
AI model inheritance (priority order):
1. Profile's AI Model (highest priority)
2. Site default model (set by Velaro admin)
3. Workflow node fallback
Common mistake: Adding an AI/Bot node without a profile set — the bot runs with no persona, uses the site default model, no greeting, and no tone. Always set a Default profile even if you only have one bot.
Troubleshooting workflows
Workflow not triggering: Check that it's Active and the trigger matches the channel the visitor is using (web, SMS, WhatsApp, etc.).
Variable is blank: Make sure a Set Data or Ask Question node ran before the node that uses the variable. Check the Execution Log to confirm.
Route to Team not assigning: Verify the team exists, has agents assigned, and the agents are online (or queue mode is enabled).
Was this article helpful?