actAVA Platform Release Notes v2 (March 2026)

actAVA Platform Release Notes

March 3, 2026

The March ’26 actAVA Release empowers organizations to advance toward becoming an Agentic Enterprise by seamlessly integrating AI agents with human expertise. This release equips every team with the insights, automation, and resources needed to confidently enhance customer experiences. It also helps administrators control the costs of their agentic solutions. 

New Features

Google Workspace Integration

Healthcare teams live inside Google Workspace — from shared care-coordination documents to scheduling and email. With this release, actAVA agents can now operate natively across Gmail, Google Calendar, Google Drive, and Google Slides, enabling workflows like automated referral follow-ups, appointment scheduling, and care-plan document generation without ever leaving the tools clinicians already know.

• 40 new agent tools spanning Gmail (send, search, label), Google Calendar (create, update, delete events), Google Drive (copy, move, organize files), and Google Slides (create and edit presentations)

• 40 new backend route handlers with corresponding service methods and 30 new Google Workspace client methods

• Simplified connector UI redesigned with per-service toggles — administrators control exactly which capabilities (Gmail send, Calendar delete, etc.) are available to each agent

• OAuth improvements, including domain validation and user-friendly error messages

• Safety-first design — Google Drive delete operations removed from agent tooling to reduce risk surface in production and demo environments

• Graceful token refresh — expired OAuth tokens are handled cleanly, eliminating 500-type errors

• 217 new tests (133 agent-side, 84 backend-side) ensuring enterprise-grade reliability

Slack Integration & Human-in-the-Loop Notifications

Modern care teams coordinate in real time through Slack. This release introduces full Slack integration with a purpose-built human-in-the-loop notification system, allowing agents to pause critical workflows and request clinician approval directly in Slack — ensuring that AI speed never outpaces human judgment.

• Full Slack OAuth flow with backend routes, callback handling, and role-based access in the Manage Connectors admin page

• Slack messaging tools have been added to the Cortex agent for sending messages and joining channels

• HITL Slack notifications — agents can notify designated Slack channels when a workflow requires human review or approval

• Slack interaction handling with internal auth for secure agent-to-Slack communication

• HITL notification tracking with a new data model and migration for full auditability

Document Themes & Export

Consistent, branded document output builds trust with patients, payers, and partners. Document Themes (formerly "Format Cards") have been elevated with a redesigned editor, expanded font support, and polished export capabilities across PDF, DOCX, and PPTX — so every artifact an agent produces looks like it came from your organization, not a machine.

• Renamed "Format Cards" to "Document Themes" across the entire UI for clarity

• Redesigned editor with config import/export, active badge indicators, and reordered sections

• Google Fonts expansion — more font family options with proper CSS stacks

• Compact preview redesign with larger preview scale and realistic sample content, including optional Table of Contents

• Multi-format preview — preview Document Theme output in DOCX and PPTX before exporting

• PDF export polish — configurable font family, cleaned callout labels, and proper table colors

• Table of Contents, anchor links, and colored indicators in DOCX export

• Organization scoping added to API calls, ensuring themes are managed at the enterprise level

AI Brand Extractor

Setting up a new agent's visual identity should take seconds, not hours. The AI Brand Extractor automatically reads your organization's PDF materials — logos, color palettes, typography, tone of voice — and generates a brand identity configuration that feeds directly into agent setup.

• New endpoint and UI modal for uploading brand PDFs

• Automated brand extraction via pypdf, pulling colors, fonts, and identity elements

• Direct integration into agent configuration — extracted brand context is applied automatically

Benchmark Usage & Cost Breakdown

Healthcare organizations need full transparency into the cost of their agentic solutions. The new Benchmark Usage dashboard provides a category-level cost breakdown for every agent run, enabling administrators to understand exactly where resources are being consumed and optimize accordingly.

• Category cost breakdown queries in the agent run repository

• Service-layer integration wiring cost data through to the API

• New "Benchmark Usage" admin tab with visual cost breakdown by category

Agent Studio Improvements

Agent Import & Upgrade

Deploying and versioning agents across environments is now seamless. Import agent definitions with a new force_upgrade parameter that validates compatibility, performs atomic upgrades, and promotes new versions — eliminating conflict errors and manual rework.

• Import Agent button in the sidebar for export/import of agent configurations

• Force upgrade with resource compatibility validation and atomic upgrade flow

• Separate MCP vs. skill validation in import compatibility checks

• User-friendly error messages on validation failures

Mandatory Core Tools

Every agent now ships with essential file system and shell tools out of the box — Glob, Grep, Read, Edit, Write, and Bash are automatically injected. No more missing capabilities during demos or production runs.

Python Code Editor

A new Monaco-based Python Code Editor brings syntax highlighting and code-completion directly into Agent Studio, making it easier to configure custom logic, evaluation scripts, and agent behaviors.

URL Deep-Linking for Run History

Run sessions are now URL-addressable — share a direct link to any specific session or run for faster debugging, team collaboration, and audit trails.

Agent Name Branding

The reserved "actAVA" prefix is now enforced for internal organization agent names, preventing naming collisions and reinforcing brand consistency across the agent library.

Evaluations & Experiments

The platform's evaluation capabilities continue to mature with improvements designed for enterprise-scale testing and compliance validation.

• Item-level rubric score calculations and filtering in the experiments UI

• Dedicated evals session type that disables HITL for eval and synthesis runs, ensuring clean automated testing

• S3 workspace fallback for multi-container AWS deployments

• Trace span reconstitution for full observability of evaluation runs

• Synthesis improvements and meta-agent custom tool handling

• Source type filtering for more targeted evaluation workflows

Infrastructure & Reliability

• S3 fallback for workspace files — files now survive container restarts with atomic write cleanup and proper logging, eliminating demo failures

• Compact streaming events before DB persistence, achieving ~40% event count reduction

• Optimized storage — ~50% reduction in run_json size and ~30% reduction in trace_span size

• Request ID + Datadog trace context propagated end-to-end through the SQS worker pipeline

• Datadog APM initialization shared across services with new monitors and threshold documentation

• User timezone support — browser timezone captured and forwarded to agent sessions

• Duplicate user profile fix — deduplication of integration connections by Google email

• Lambda invoke retry logic improvements

• Staging releases v17–v20 promoted and verified

Design System & Developer Experience

• Formalized style guide and color token governance with lint enforcement for hardcoded colors

• Dark mode fixes, including Safari prefix and warm dark muted-foreground adjustments

• Standardized typography across markdown renderers

• Account settings page and settings layout wrapper with the Skills tab restricted to system admins

• System admin org-scoped access — view any organization's usage via X-Organization-Id header

• Confirmation dialog before creating new agent versions

• Active run highlighting and state management race condition fixes