Chapter 2 — Studio(Tutorial example at the end)
Why this chapter matters
Studio is where you prepare everything: forms, message templates, automated workflows, data sources, AI personas, and CUI Apps Package. Day-to-day operators usually run published flows from Chapter 1 Apps.
This chapter follows the Studio section of the left menu, top to bottom: Form management → Message templates → Workflow designer → Dataset management → AI staff profiles → CUI Apps Package.
Examples:
- A trader connects a weekly price sheet in Dataset management so a quoting flow always reads the latest unit price.
- HR builds an “offboarding checklist” in Form management, then uses Workflow designer to send it for completion.
- Marketing maintains approved opening templates in Message templates so support flows can legally message new customers proactively.
2.1 Form management
Design forms for external or internal use (fields, layout, required or not). In a flow, the “Send e-Form” step sends a link; after submit, managers approve under Apps → Running forms.
For general web forms (HTML), the designer includes an AI assistant to help generate or adjust fields and layout from your description; exact placement and steps follow the product UI.
Two form types (how to choose)
| Type | Best for |
|---|---|
| Standard web form (HTML) | Flexible layout, or sharing a link to fill in a browser (quotes, event signup, etc.). |
| In-WhatsApp form (Meta Flows) | Customers fill inside WhatsApp chat; after Meta review, works better with template messages. |
- Click Add and pick a form type.
- Use the designer to add fields, required flags, and help text, then save.
- Activate the form so it can be selected in the “Send e-Form” workflow step.
Form type is fixed at creation; you cannot convert directly. If you need both, create two forms.
2.2 Message templates
Central place for WhatsApp message templates (including Meta-submitted templates). To “send a template message” proactively in a flow, a usable template must appear here first.
Important: WhatsApp connection credentials for the official channel are set when editing the company under Admin → Company / user management — not the same as the API management page.
- In Workflow designer, open a flow and add a “Send Message” or “Send template” style step.
- Pick a template and map flow data to variable placeholders (e.g. customer name, order id).
- Save the flow; at runtime the system fills those values.
2.3 Workflow designer
Visually chain “who gets what, when, and whether to wait for a reply or continue automatically” into an automation playbook. After activation, manual flows appear under Apps → Published apps.
The Workflow designer includes an AI assistant to help generate or adjust nodes and connections; the exact entry point and steps follow the product UI.
How the designer layout works
- Left: node palette grouped by category.
- Center: canvas; edges show order between nodes.
- Right: properties for the selected node.
- Process variables: available from the toolbar — per-run scratch data (e.g. last customer reply) for later steps; exact names follow the UI.
Common list actions
- Add: create a new flow and open the designer.
- Edit (design): change nodes and edges.
- Manual start: run one instance immediately for testing.
- Activate / deactivate: deactivated flows cannot be started from Published apps; auto triggers stop too.
- Copy / delete: copy keeps structure under a new name.
On the start node, choose whether colleagues start the flow manually from Published apps or the system starts it on events such as inbound messages. One start node per flow.
Advanced (per designer options): the start node can also create runs on a schedule or dataset condition, or when the inbox receives a matching email. Mail rules often support include/exclude on sender, subject, body keywords, attachments, extensions — similar idea to the “Wait for email reply” node.
Node reference (names follow Traditional Chinese UI)
Below: each palette node’s purpose and typical use (labels may vary slightly by version — trust the screen).
(1) Start and end
| Node | What it does | Example |
|---|---|---|
| Start | Entry point; choose manual start vs automatic on message/system events. | Event signup only after a web button → manual; auto greeting for support → automatic. |
| End | Completes the run; you can wire multiple ends (e.g. “success end”, “customer declined end”). | Quote declined → one end; deal closed → another end. |
(2) Messaging & commerce
| Node | What it does | Example |
|---|---|---|
| Send Message | Send WhatsApp text or media; template mode may be available per UI. | Poster image plus caption; personalised greeting with customer name. |
| Send template | Sends an approved template message for proactive outreach. | After shipping, send “Your order has shipped” with tracking id. |
| Institution ordering cart | Guides “browse categories → pick items → adjust qty → cart → submit” ordering in chat (with WhatsApp Flows). | School lunch programmes, B2B batch orders inside chat. |
| WhatsApp cart | Legacy multi-product cart; may be hidden by default in the palette — only older flows may still use it. | If your consultant migrated you, you might still see it; new designs usually use institution ordering cart. |
| AI chat ordering cart | An AI employee chats with the customer in natural language. The customer types or sends a product image; the AI matches items from the catalogue, builds a cart, and asks for confirmation. On confirm, a Sales Order is created automatically — no button-by-button selection needed. | A restaurant customer types "two rice boxes and one fish ball noodles" and the AI assembles and confirms the order; a B2B buyer sends a photo of a purchase list and the AI creates the order. |
| AI chat bot | An AI employee conducts open-ended multi-turn conversations, drawing from a knowledge base (datasets or Studio documents). Unlike the chat ordering cart, there is no order creation — pure Q&A and guidance. The flow continues when the customer says a configured "end conversation" phrase. | Customer service FAQ bot (product specs, return policy); automated onboarding Q&A for new employees; pre-event info chatbot. |
(3) Waiting for customers or external input
| Node | What it does | Example |
|---|---|---|
| Wait for reply | Flow pauses until the customer sends a WhatsApp message; optional prompt, who may reply, format checks or AI validation. If the UI offers knowledge base / dataset / studio documents, internal reference can guide AI validation (same idea on “Request location” and similar nodes). | Ask “Plan A or B?” and branch on the answer. |
| Wait for QR code scan | Customer uploads an image containing a QR code; decoded value is stored for later steps. | Event check-in, voucher redemption, scanning a standee code. |
| Request location | Optionally send a template first, then ask the customer to share location — useful for delivery or field service. | Confirm drop-off area; rally point for events. |
| Wait for email reply | Pause until an email matches rules (sender, subject, body keywords, etc.). | When a vendor replies “confirm shipment”, internal steps continue automatically. |
(4) Conditions & data
| Node | What it does | Example |
|---|---|---|
| Conditional branch | Split on prior data: if condition → path A, else path B (or default). | Route “refund” mentions to a specialist queue; everything else to general Q&A. |
| Dataset query | Query, insert, update, or delete against datasets you built in Dataset management (per node settings). | Look up contract price by customer id; write an order row back to a sheet. |
(5) Integrations
| Node | What it does | Example |
|---|---|---|
| Call external API | Issues an HTTP request to other software (orders, CRM, etc.) — usually configured by IT or a partner first. | After order confirmation, create a sales order in ERP. |
| Calendar | The AI reads conversation text, dataset rows, form submissions, or email attachments and generates calendar commands (create / update / delete events), which the system executes immediately. The AI checks existing events first to avoid double-booking. | Customer describes a meeting request in chat — AI creates the calendar entry; a contract PDF is parsed and expanded into monthly reminder events. |
| Odoo integration | Creates or queries documents in a connected Odoo 19 ERP instance: purchase orders, quotations, sales orders, invoices, deliveries, goods receipts, returns, credit notes, payments, and partner / product / stock lookups. AI can extract fields from PDFs, images, or conversation text automatically. | WhatsApp order confirmed → Odoo sales order created and delivery dispatched; supplier sends a PDF invoice → AI extracts line items and amounts into a purchase order. |
(6) Forms & documents
| Node | What it does | Example |
|---|---|---|
| Send e-Form | Creates a form link and sends it; after submit, managers approve under Running forms. | CSAT survey, goods receipt checklist, supplier qualification form. |
| Generate HTML report | Uses AI with your brief and data to build a printable web report. | Weekly complaint digest for leadership; meeting minutes layout from chat. |
| Generate document | Uses AI to produce structured content, then Word / Excel / PowerPoint files for download or sending. | Auto-filled quote draft, contract draft, stocktake sheet. |
Publish a flow from scratch
- Workflow designer → Add, drag Start, middle nodes, and End, connect with edges.
- Select each node and complete the right panel; open process variables from the toolbar if needed.
- Save, then activate the flow in the list.
- Manual flows show up under Apps → Published apps.
2.4 Dataset management
Connect tabular company data (inventory, spreadsheets) so flows can query or write back without copy-paste. Pairs with the Dataset query node.
- Click Add, name the dataset clearly, pick a source type (per UI).
- Follow the wizard to connect the file or source and verify column names.
- Use preview or test, confirm data looks right, then save.
- The dataset becomes selectable in Dataset query nodes inside Workflow designer.
If sync fails, the UI usually explains why; fix the source or connection and retry.
2.5 AI staff profiles
Define personas for virtual assistants: display name, avatar, role text, notes — so customer-facing or internal AI tone stays consistent. If the menu item is missing, your role does not include this feature.
Click Add, complete each tab in the UI, then save. If the avatar preview fails, try re-uploading or confirm you are signed in.
2.6 CUI Apps Package
Bundle a workflow plus related forms, templates, etc. into one file for copying to a test tenant, another company, or backup/restore — common when a partner delivers a standard CUI Apps Package.
- Export: pick a workflow from the dropdown → run Preview to see bundled resources (forms, datasets, message templates, Meta templates, AI staff, knowledge documents, etc. — as shown) → uncheck anything to omit → start generation. Generation runs in the background; when done, download the
.studio-bundle.ziparchive. Filenames typically include the workflow name and a timestamp. - Import: upload the same bundle format → preview what will be written → if names collide, the UI lets you set import-as names → confirm import (may also be asynchronous — follow progress messages).
Preview completeness depends on bundle contents and API permissions; if preview looks wrong, check roles and source environment version with your admin or consultant.
TutorialExample: an “AI HR / expense assistant” style flow (diagram 1)
Goal: colleagues request leave or upload receipts on WhatsApp; the system checks company rules, uses AI where needed, then routes items needing approval to managers.
- Form management: leave, expense, or onboarding forms (fields per policy).
- Message templates: proactive or follow-up templates (requires WhatsApp company connection and Meta approval).
- Dataset management: leave policies, GL accounts, etc. for Dataset query nodes.
- AI staff profiles: e.g. “HR assistant”, “Expense reviewer” personas for nodes to reference.
- Workflow designer: typical chain — “Send Message / template” → “Wait for reply” (validation + AI staff) → “Conditional branch” → “Send e-Form” or “Dataset query/update” → optionally “Generate HTML report / document” → “End”.
How the product ties together: “Wait for reply” can bind validation to an AI staff member; “Generate HTML report / document” can pick an AI provider and staff, and optionally use datasets as knowledge (matches backend flow spec). Field-level UI depends on your build.