The March 2026 edition brings major updates across the platform — including Guardrails for AI Colleagues with built-in PII detection, content moderation, and jailbreak prevention; enhanced Case Management skills for full ticket lifecycle handling in chat; Skill Callbacks for async workflow resumption; and a new visual Task Journey builder for onboarding. On the Helpdesk side, there's bulk ticket status updates, customizable email branding, audit logs, and a rich text editor. Platform updates include integration alert rules, Workday RaaS support, user-level Workday authentication, web app performance and resilience improvements, spam detection on phone channels, and several security and stability fixes.

In This Edition

---

AI Colleagues Studio

A significant set of updates this month — from enterprise-grade safety guardrails and richer case management capabilities to async skill execution, smarter data handling, and voice channel improvements.

Guardrails — Centralised Safety Controls for AI Colleagues

Admins now have a centralised safety layer that runs on every AI Colleague execution. Configurable from Settings → Orchestrator Settings → Guardrails, three types of guardrails are available:

PII Detection — Catches personally identifiable information in both user inputs and AI outputs. Select the entity types relevant to your organisation (Aadhaar, SSN, email, phone, credit card, NHS, and more) and choose the action on detection: Mask & Continue (default), Block, or Log & Continue.

Content Moderation — Detects harmful content categories (sexual, hate, harassment, self-harm, violence, illicit) across inputs and outputs. Action is always Block.

Jailbreak Detection — Identifies attempts to override instructions or manipulate the AI outside its configured scope (role hijacking, prompt injection, instruction override patterns). Runs on user inputs only. Action is always Block.

📖 Documentation: Guardrails

Case Management Skill Enhancements

The Case Management skill has expanded well beyond basic ticket creation and status lookup. Employees can now handle the full ticket lifecycle directly in chat:

Raise on behalf — Employees can raise tickets for a colleague directly through the AI Colleague. Smarter form filling — Fields auto-prefill from employee profile and conversation context, with adaptive handling based on form structure.

Attachment support — Files can be attached during ticket creation and when adding comments.

Full ticket visibility — "Show me my tickets" now returns all relevant tickets — those raised by the employee, raised for them, and raised on behalf of a colleague.

Richer ticket details — Fetching a ticket now returns the full picture: public comments, approval status, pending-on-user flag, closure date, watchers, and all custom form fields.

Add comment in chat — Employees can comment on existing tickets directly in the conversation, with optional file attachment.

Increased form and category limits — Form limits increased from 10 to 20, and the category limit of 100 can be removed entirely by enabling Large Data Handling. All enterprise support scenarios are now reachable.

📖 Documentation: Case Management Skill

Skill Callbacks — Async Workflow Resumption

AOPs can now pause mid-execution and wait — for a long-running or asynchronous skill to return a result, or for a user to submit a form shared during the flow. Execution resumes automatically once the response comes in. This is powered by a webhook-based callback mechanism with built-in timeout handling and automatic retry, ensuring reliable resumption even for workflows that take minutes to complete.

📖 Documentation: Understanding Skill Execution Scope

Large Data Handling

AOPs can now process large, unpaginated responses intelligently. When an API or connector returns a high volume of data, a dedicated agent kicks in automatically, embeds the user's query and each returned item, and uses similarity matching to surface only the most relevant results — so users get a clean, accurate answer instead of an overwhelming or broken experience. This capability is currently being rolled out selectively and is flag-driven for evaluation.

AOP Creator — Bypass Validation

You can now skip validations when publishing an AOP. If you're iterating or testing, you're no longer blocked by warnings before hitting publish — giving you more control over when and how you ship changes.

Global Search

Search now works across AI Colleagues, AOPs, Workbench, and Skills — all from one place. No more switching contexts to find what you're looking for.

Voice Improvements

Several voice channel enhancements this month:

• Context continuity across voice and text — The orchestrator now maintains context between voice and text channels, so users can seamlessly switch without losing conversation history.
• Configurable small talk frequency — Admins can now control the frequency of filler messages between actual voice responses.
• Additional instructions support — Voice orchestrator behaviour can now be fine-tuned with additional instructions.

📖 Documentation: Voice

UX Improvements & Bug Fixes

• Dynamic dropdown skills (API/connector-driven) now work correctly when helper or primary AOPs are triggered by the system.
• Helper AOP runs initiated from a primary AOP are now grouped under the same run ID.
• For-each loops: Drill-down now supported for helper AOPs inside for-each loops, and multiple processing issues resolved.
• Status chip inconsistency across AIC Studio has been resolved.

---

Helpdesk

Customizable Dashboard Report Configuration

Dashboard reports are now configurable at the time of download. Users can add or remove columns and reorder them based on preference. Configuration applies across all selected filters. Previously, dashboard reports had a fixed column structure — now users can tailor downloaded reports to their needs, making them easier to analyse and share.

📖 Documentation: Report Configuration

Instant Trigger for Custom Reports

Custom reports can now be triggered instantly via email. Earlier, reports could only be scheduled — users can now send the report immediately without waiting for the next scheduled run.

Improved Ticket PDF Download

Ticket PDFs now have a clean, structured layout for better readability, with improved spacing, formatting, and clear separation of ticket details and content. Makes exported tickets easier to read and share for documentation or audits.

Bulk Status Update

Users can now update ticket status in bulk. Perform status changes on multiple tickets at once for tickets under the same ticket schema — reducing manual effort for large ticket volumes.

📖 Documentation: Bulk Status Update

Email Notifications — Custom Branding & Preferences

Admins can now customise and control email notifications more effectively:

Custom Branding — Add header and footer images to emails, configure a custom footer message, and ensure all outgoing emails align with your organisation's branding.

Email Preferences — Set up role-based notification rules and control which events trigger emails and who receives them. Helps reduce unnecessary notifications and improve relevance.

📖 Documentation: Email Notification

Setting Logs (Audit Logs)

Helpdesk events are now available in Audit Logs. Track what change was made, when, and by whom — covering key admin setting updates and actions. Improves visibility and traceability of changes for auditing and debugging.

📖 Documentation: Audit Logs

Rich Text Editor

A rich text editor is now available in ticket descriptions and comments across Dashboard and Webview. Supports formatted text for better readability, and is available while creating and updating tickets.

---

Integrations & Connectors

Integration Failure Alerts

Admins can now set up automated alert rules for integration failures. Under Settings → Integrations → Alerts, you can define condition-based rules that monitor both custom integration logs and connector execution logs, filtering on fields like request URL, status code, app name, action, and execution status. Rules support nested AND/OR logic, configurable frequency (occurrences + time window + cool-off period), and email notifications to all or specific integration admins.

📖 Documentation: Alerts

Workday RaaS (Report as a Service)

A new action — Run Custom Report (RaaS) — is now available under the Workday REST connector. Admins can configure the report owner, report name, and pass dynamic input parameters as key-value pairs. The system constructs the standard Workday RaaS endpoint and returns the report data as JSON. This enables AI Colleagues to pull live Workday report data as part of any workflow or conversation.

📖 Documentation: Workday

Workday User-Level Authentication

End users can now authenticate directly with their own Workday credentials via OAuth 2.0 to access personal data like approvals, pay slips, and time-off. This is distinct from the system-level Integration System User (ISU) auth used for background syncs. The flow redirects the user to Workday's authorization page, obtains a user-scoped token, and securely stores the refresh token for subsequent requests — enabling personalised actions without exposing admin-level credentials.

📖 Documentation: Workday

---

Onboarding

Task Journey — Visual Flow Builder for Onboarding

Task Journey lets you visually design and automate the sequence of onboarding tasks using a drag-and-drop canvas. Drag tasks onto the canvas and connect them to define execution order — including parallel and converging paths. Schedule tasks relative to Date of Joining or Last Working Day with day, week, hour, or minute offsets. Add conditional logic so one journey adapts to different employee segments (department, location, etc.). Built-in journey validation catches misconfigurations before publish.

📖 Documentation: Task Journey

---

Workflows Studio

Automatic Task Reassignment on Employee Termination

When an assignee is terminated, pending tasks can now be automatically reassigned — ensuring nothing falls through the cracks during offboarding transitions.

Application-level toggle — Enable or disable auto-reassignment per workflow app via the "Automatically reassign tasks on assignee/s termination" switch in the Create/Edit Application form.

Custom notifications — Configure dedicated auto-reassignment notification templates (separate from manual reassignment) in the Notifications tab of each Approval/Input node.

The system automatically detects terminated assignees, re-evaluates assignment rules against your current HRIS data, and falls back to the configured default assignee if needed.

📖 Documentation: Auto-Reassignment on Termination

---

Platform & Global Settings

A broad set of updates spanning web performance, network resilience, channel experience, security, and overall platform stability.

Web Experience & Performance

Lazy Loading for Heavy UI Modules — Conditional UI modules (richer media, thread/voice components) now load on demand, resulting in faster initial page loads and lower memory usage.

Accessibility Monitoring — On approximately 10% of page loads, the app now runs WCAG 2.1 accessibility checks on core screens, sending results to PostHog for ongoing visibility into accessibility issues without impacting users. Current score: 93.74.

Better Behaviour on Unstable Internet — The app now handles poor connectivity more gracefully with fewer disruptive pop-ups during connection drops, automatic message sending after reconnection, and clearer "No internet" pages with a branded experience and Retry action.

Mark All as Read — Users can clear all unread notifications with a single click.

Channels

Cleaner Slack Experience — Slack responses are now cleaner (no raw technical code shown), key disclaimers display reliably, and the app manifest now includes assistant permissions and events for assistant-style flows. Messages with headers are converted to Slack-friendly formatting, and tables render with Block Kit.

Microsoft Teams Retry Logic — When notification delivery fails with "Conversation not found", the system schedules up to 3 retries (30s, 90s, 180s) and may refresh conversation context in the background, reducing silently lost notifications.

Safer Phone Channel with Spam Detection — Added spam detection to filter unwanted or abusive calls, protecting users and agents.

More Resilient Slack Event Handling — Slack delivery events are now processed via a queue (rather than all at once), preventing overload when large broadcasts trigger a surge of acknowledgement events.

More Reliable Voice Input — Improved voice call error tracking for network-related issues and fixed an issue where voice input could stop working in Safari after switching tabs.

Security & Infrastructure

SAML Certificate Expiry Alerts — Automated checks detect SAML certificate expiry within 15 days and send grouped reminder emails to Config Manager users, reducing the risk of unexpected SSO outages.

Stricter AI Safety Checks — Enabled stricter compile-time checks for AI components to prevent certain categories of errors early.

More Robust Authentication Flows — Improved login token validation and safer handling of auth-related data for a more stable single sign-on experience.

Safer Editing of User Data — Fixed a critical issue where deleting a user row could sometimes delete the wrong one.

Stronger User Sync — Improved how user records are synchronised across systems so user counts stay accurate and the sync process scales better.

Audience-Specific Branding — Branding and configuration can now be tailored for specific employee audiences, enabling different experiences for different groups without manual workarounds.

📖 Documentation: Audience-Specific Branding

Granular Template Access Control — Added more granular access control for communication templates so ownership and editing can be safely delegated.

📖 Documentation: Letter Templates

Bug Fixes & Stability

• Improved conversation stability for long threads — fixed an issue where long conversations could lose earlier messages
• Corrected a routing issue where some users were redirected to outdated environments after signing in
• Fixed issues with users being wrongly terminated on some channels
• Resolved empty messages being sent in Workvivo and Teams navigation instability during voice calls
• Better handling of complex web links with spaces or extra parameters in embedded views
• Smarter handling when AI services aren't configured — the app now stops making failing health-check calls in environments where AI is off by design