Skip to main content

​
March 5, 2026

​
Agents & AI

Agent Instructions Version History - Users can safely iterate on agent instructions freely, track what changed, and roll back instantly if something breaks.
  • A history of instruction changes on the agent.
  • Side-by-side version comparison for any previous agent for Admins.
  • Reverting to a previous version saves it as the new current version.
Agent Conversation History Waterfall View - Get clear visibility into how a conversation executes, making it much faster to debug and optimize agent performance.
  • Agent admins can view a waterfall breakdown of all requests in a conversation — tools, responses, and timing in a single view.
  • Each step shows an estimated duration so it’s easy to spot where time is being spent.
  • Hovering over any step previews the message without leaving the waterfall view.
Agent Skills - Agents can now discover and execute Skills at runtime — a new way to build modular, reusable intelligence into your apps.
  • Create and manage Skills within an App, each with a name, description, prompt instructions, and configured automation tools.
  • Admins control which skills an agent can access: all, none, or a specific selection.
  • Agents can automatically discover and invoke the right skill for the task based on the name and description — no additional setup needed.
Multi-Modal on Gemini-Powered Agents - When using Gemini in the Admin Builder, Personal Assistant or Record Details, attach images or PDFs with your message to the agent for extra context. AI Search Table Creation Enhancements - Admins gain better insight into search table setup, including setup status and improved recovery capabilities.
  • AI Search tables are now created immediately in the Element, with a status indicator showing whether the underlying table is ready.
  • Admins can reference and configure a search table right away, but it won’t be functional until it reaches a ready state.
  • In case of error, admins can re-try the AI search creation with one click.
When an admin is configuring an AI search tool within an agent, they can specify input value references with descriptions for how the agent can use them.
  • Input values can then be referenced in the filters.
  • An agent can pass values into the input parameters.
AI Search from Automations - Admins can add an AI Search action step within an Automation to bring semantic search directly into your workflow.
  • Select any configured AI Search table as the search source.
  • Add a dynamic query using value references from earlier in the automation.

​
Automations

Webhook Trigger Custom Response Bodies - Define custom response bodies on a webhook trigger to integrate with external systems with a specific response format when calling your webhook endpoint.
  • Custom Response field when creating or editing a Webhook trigger.
  • Formats include JSON, DML, or TEXT.

​
Reporting

Report Layout Configuration Flexibility - When generating PDF reports, use custom formatting to efficiently share and analyze the desired information.
  • Choose to display associated reference values for dynamic dropdowns.
  • Merge cells to create headers or consolidate information.
  • Apply text formatting to reference fields, including bold, italic, font sizing, and styles.
  • Duplicate PDF Report Templates to easily share formatting between several reports.