v2026.4.1
Release date: April 1, 2026
✨ New Features
AI Agent tools for customer profile data
New system actions let the model update customer record fields during a conversation, with server-side validation.
What is it for?
- Change Customer Email — saves the email the customer provides in the chat.
- Change Customer Document — saves tax ID (e.g. CPF/CNPJ in Brazil) according to country and person type (configured defaults or context). Values are stored as digits only (no punctuation).
- Change Customer Country — updates the country code (e.g. ISO alpha-2).
- Change Customer Person Type — sets natural person (
individual) or legal entity (company).
For the document tool, the API schema requires document, country, and person_type; admins set defaults in the editor, and the model may override them when the conversation context calls for it.
How to use?
- Open AI Agent from the sidebar and edit the desired agent.
- Go to the Tools tab.
- Click + Add System Action.
- In Customer & data or Service & handoff, pick the action (and configure country/person type for the document action when needed).
- Save the agent.
System actions modal grouped by section
The modal that lists system actions now groups options by theme for easier scanning.
Sections
- Customer & data — name, custom fields, email, document, country, person type.
- Service & handoff — transfer to team.
- Scheduling — schedule and confirm appointment.
- Flow & funnel — change funnel and return date.
- AI agent — unknown response fallback.
Sections with no visible items (e.g. unique actions already added) are hidden.
System context includes email, document, and country
In the initial context message to the model (prepareContextMessages), the Customer info JSON block now includes email, document, and country when a customer_id is present, in addition to the name.
🎯 Benefits
- ✅ Fewer handoffs just to update basic customer data
- ✅ Tax IDs stored in a consistent format (digits only) for integrations and reporting
- ✅ Clearer agent editor with section grouping
- ✅ The model sees richer registered customer data at the start of the thread