Why you need workflows
Every job has repetition. Checking inventory each week, finding low-stock products, comparing suppliers, creating purchase orders, getting approvals, and reporting results. It takes one hour each time, but done weekly, that adds up to 52 hours a year.
Traditional automation tools tried to solve this problem, but they required development skills or complex configurations. Even with drag-and-drop block connectors, you still had to manually set up API authentication, data transformations, conditional branches, and loop handling.
wiiiv workflows are different. Describe what you want to do through conversation, and AI designs an executable workflow. From login to purchase order approval and email delivery, AI assembles every step, and all you need to do is confirm.
How it differs from existing workflow tools
Limitations of existing tools
Tools like Zapier, Make, and n8n use a "trigger + action" model. Users manually select each step, map fields, and configure authentication. Once conditional branches or loop handling are added, complexity increases dramatically.
The wiiiv approach
- Conversational input: Just say "Look up SkyMall best sellers and low-stock items, check supplier performance, place orders, and send an email report"
- AI design: The entire flow is assembled automatically: authentication, API calls, data joins, loops (supplier lookup / performance check / order creation / approval), report generation, and email delivery
- Governed execution: Every step leaves an execution record and verification trail, providing transparent tracking of what was executed and how
- Reuse: Once created, a workflow can be re-run with a single button, shared with teammates, or called externally via API
How to read a node map
When you create a workflow, a node map is generated. Each node represents a single task step, and they execute sequentially from top to bottom. Below is the node map for an actual "Supplier Order Audit and Email Report" workflow.
if all-have-order-history → proceed to create_normal_notice step
Node map header bar
- View toggle: Switch between Node Map (graph), Detail List (text), and Context (design rationale)
- API tag: Indicates whether this workflow calls external APIs
- Service badges: External systems integrated (SkyMall, SkyStock, Report, Email, etc.)
- Hash and node count: Unique workflow identifier (#f23323ef) and total number of nodes (10 nodes)
Node card structure
Each node card displays a number (#1), service icon, node name, type badge (ACT, COMMAND, DECIDE, etc.), and service badge (SkyMall, Email, etc.). The body contains a description of what the step does, and the footer shows input/output variables.
Clicking a node flips it over to reveal technical details such as API URL, headers, body, and hints.
Node types
wiiiv workflows consist of 7 node types. Each type is distinguished by its background color and left-side color line.
In addition to the above, there are Input (user input) and Command (Python script execution) types. Command is used when AI generates and runs Python code for joining multiple data sources or performing complex calculations.
Connections between nodes
Nodes are connected through variables. The Output of one node becomes the Input of the next. For example, the Output best_sellers from #1 fetch-skymall-best-sellers is passed as Input to #4 select-target-products.
Two execution modes
wiiiv workflows can run in two modes. Choose from "Workflow Execution Mode" in the settings screen.
- Each node is converted to a Python script and executed directly
- Runs without LLM calls, so execution speed is fast
- Guarantees the same result for the same input every time
- Calls LLM only minimally where judgment is needed (
llm.decide(),llm.generate()) - Suitable for most workflows
- The HLW engine calls the LLM for each node during execution
- Provides detailed interpretation and logs for execution results
- Handles unexpected situations flexibly
- Suitable for tasks requiring judgment, such as report writing and analysis
- Takes more execution time and cost than P Line
Creating workflows from chat
The most natural way to create a workflow is to describe the task you want in a chat.

skymall-bestseller-lowstock-procurement-and-email
v1.0 · 13 nodes
Looks up SkyMall's top 20 best sellers and products with 30 or fewer units in stock, queries SkyStock for CRITICAL inventory alerts, then filters out products that are selling well but low on stock. Checks suppliers and their performance for those products. Creates and approves purchase orders for qualifying products, then sends the full results via email.
Once AI generates the node map, you can review it right inside the chat. Say "run it" to save and execute simultaneously.
Tips for writing good requests
- Be specific: Instead of "look up products," say "look up products with 30 or fewer units in stock from SkyMall"
- Include outcomes: Describe what to do after retrieval (report, email, order placement)
- State conditions: Provide judgment criteria, such as "products selling well but low on stock"
- Multiple systems: When integrating several systems, specify what each one should do ("look up from SkyMall, and also check SkyStock")
Running a workflow
1. Run from the Workflows page
Select a workflow from the Workflows page, then press the ▶ Execute button in the top right. If there is an Input node, a screen for entering values appears first.
2. Run from chat
Say "run the skymall-bestseller-lowstock workflow" and it executes right inside the conversation.
3. Run via API
Check the endpoint and curl example in the API tab on the workflow detail screen.
4. Scheduled execution (CRON)
Set up a recurring execution schedule in the CRON tab on the workflow detail screen. You can enter a CRON expression directly, or type conversationally like "every Monday at 9 AM" and it will be converted automatically.
Workflow list and management
All workflows in a project are listed on the left side of the Workflows page.
Filters
- All: All workflows in the project
- Mine: Only workflows I created
- Shared: Only workflows shared with teammates
- Auto: Only workflows automatically generated by AI from chat
Right-click menu
Right-clicking an item in the workflow list reveals the following actions.
- Edit in chat: Navigate to the chat screen and modify the workflow through conversation
- Execute: Run immediately
- Duplicate: Create a copy of the workflow
- Rename: Change the workflow name
- Assign color: Set a color tag for categorization (7 colors)
- Toggle sharing: Switch between PRIVATE and PROJECT (team shared)
- Export: Download as a .hlw.json file
- Delete: Permanently delete (confirmation dialog shown)
Editing a workflow
The way to modify an existing workflow is to edit it in chat. Select "Edit in chat" from the context menu, and a chat session with the workflow loaded will open.

Changes:
• #2 fetch-skymall-low-stock: threshold 30 → 20
• #13 send-email-report → send-slack-alert: Email → Slack #inventory-alerts
• Service badge: Email → Slack
The remaining 11 nodes are unchanged. Shall I apply?
When you request changes through conversation, AI modifies only the necessary nodes and keeps the rest intact. You can also request multiple changes at once.
Reuse and extension
SubWorkflow calls
If you save a frequently used task (e.g., SkyMall authentication + token extraction) as a separate workflow, other workflows can call it via a SubWorkflow node.
Duplicate and modify
Duplicate an existing workflow, then modify only a few parts in chat to quickly create a variation.
API integration
Every workflow has a unique API endpoint. External systems or scripts can call workflows programmatically.
Sharing
Switching from PRIVATE to PROJECT allows teammates in the same project to run the workflow.
Upload / Export
You can export and upload as .hlw.json files, enabling workflow transfers between projects.
Practical tips
Workflow names
AI generates names automatically, but a target-action-result pattern makes them easier to find later. e.g. skymall-bestseller-lowstock-procurement-and-email
onError policy
Each node has a failure behavior configured: retry:2 (retry twice), skip (skip), abort (stop). AI sets these automatically based on each node's nature, but you can change them in chat.
Node map zoom
Use the - 100% + controls in the top right of the node map to adjust zoom. The minimap in the bottom right provides an overview of the entire structure.
Context view
Press the 💡 (lightbulb) tab at the top of the node map to view the Genesis Context, where AI explains why it designed the workflow this way. It records the purpose, key decisions, and things to watch out for when modifying.