TL;DR: Agents can now start approval flows (with user-chip approvers), send email through Gmail and Outlook, and act on the sub-items of a newly created project or folder for blueprint kickoff workflows. Plus: a limited-time free-access code for the Wrike AI Elite Mastery certification ($299 value) details below.
Featured example: the Blueprint Kickoff Agent assigns every sub-task, sets every date, and emails the CSM the moment a project is created from a template.
Hey Community! 👋
Welcome to Wrike Agent News!
This monthly digest features updates, learnings, and how-tos from the Wrike R&D team.
This month’s theme is closing the loop: March was about giving agents finer control over what they see and where they act. April is about letting them finish the job: start approvals, notify people outside Wrike, and set up an entire project from a single trigger. Workflows that used to end at “agent flagged it” can now end at “agent handled it.”
🎓 Wrike AI Elite Mastery Certification
We’re offering readers of this newsletter free access to the Wrike AI Elite Mastery certification: a $299 program covering agent configuration, prompt design, and real-world use cases, with a recognized completion credential.
Redeem with voucher code: BUILDER
Valid until: May 30, 2026
Use the code to register here. The code is single-use per person but available to everyone reading this. Feel free to share it with teammates who work with agents.
Why now? As more features land, and this month is a big one with approvals, email, and blueprint kickoff, there’s more to learn. If you’ve been configuring agents by trial and error, the certification shortens that loop significantly.
✅ New Action: Start and Manage Approvals
Agents can now start approval flows and manage approvers directly.
You can configure the action with user chips: specific people or user groups embedded in the prompt, or let the agent select approvers from context, such as the task author, project owner, or a person referenced in a custom field. The agent reads the task, decides whether an approval is needed and who should approve it, then starts the flow.
There are two variants:
- Start approval flow: Agent creates a new approval on the task with the right approvers.
- Manage approvers: Agent adds or removes approvers on an existing approval as context changes.
Why this matters: Most approval flows today are either manual (“someone remembers to start one”) or hardcoded (“always sent to the same person”). An agent can read the brief, recognize that this request needs legal review and a creative lead, and launch the right approval flow with the right approvers. That’s judgment, not a fixed rule.
User chips in the prompt give you a controlled shortlist. You define the candidate pool; the agent chooses the match.
📧 New Action: Send Email (Gmail and Outlook)
Agents can now send email through your Gmail or Outlook account.
Connect your account once in agent settings, then use the Send email action. The recipient address is included in the prompt, along with the subject, body, and any CC or BCC. The LLM generates the message from task context and your instructions, then sends it from your connected account.
Use cases:
- Escalation notifier: when a task reaches a critical status, the agent sends an email to the CSM with the needed context.
- External status update: a client-facing project update, personalized per project, sent from your work address.
- Cross-system handoff: the agent summarizes the task and emails an external team that does not work in Wrike.
Why email, not just comments?
Because many stakeholders do not work in Wrike. Clients, legal, finance, and vendors live in email. Until now, agents could only act inside Wrike. Now they can communicate outside it too.
Attachments, CC, and BCC are supported. Just specify them in the prompt. For example: “Include attachments,” “CC the project owner,” or “BCC the shared mailbox.” The agent reads the task context and fills in the rest.
🧩 New Action Scope: Sub-Items of the Triggered Item
Agents can now act on the sub-items of a newly created project or folder, not just the item that triggered the agent or sub-items of the appointed item.
This enables the blueprint kickoff pattern. You can trigger an agent when a new project is created from a specific blueprint, and have its actions run on each new sub-task: assigning owners, setting dates, updating priority, or changing statuses. Previously, agents could act only on the project itself, which meant separate agents or manual work for the rest.
Use cases:
- Blueprint kickoff: a new client project creates 20 tasks; the agent assigns each one based on the task title and the project’s “Client Tier” field.
- Post-creation cleanup: a new intake project is created with scaffolding tasks; the agent sets due dates on each based on project urgency.
- Template-to-active promotion: when a project moves from “Template” to “Active,” the agent fills each sub-task with project-specific context.
Combined with parent item reading (from March), a sub-task can now access its parent’s custom fields. That means the agent can set a sub-task due date based on the parent project’s start date, client tier, or category.
👥 Approvals With User Chips, Polished
Following up on the March-sprint approvals work, the user-chip experience in approval actions is now fully in place.
You can now:
- Embed specific users as approvers directly in the prompt using chips.
- Include user groups as chips.
- Mix chips with conditional logic, for example, if the task’s Risk Score is “High,” add the Legal group as an approver.
Also fixed this sprint: chips stay current when the underlying user or group is renamed, invitees can be selected, and validation catches deleted users.
Bug Fixes and Polish
A number of under-the-hood improvements went out this month, especially for teams who were impacted by edge cases or usability issues:
- Custom fields with "project only" or "project and folder" propagation are now available in triggers and actions. Previously, these fields were incorrectly filtered out.
- Agents now stop cleanly after deactivation. If an agent is turned off while it is mid-run, any actions already in progress will no longer continue executing indefinitely.
- Cloned agents in cloned accounts now correctly map to cloned assets including workflows, statuses, custom fields, and mentions, eliminating the need for manual reconfiguration after cloning.
- Translation issues in French and other locales have been fixed, including the {bStart} / {bEnd} formatting leakage that appeared in scope labels.
- Logs UX:
- sort order now persists after refresh,
- agent names truncate properly when space is limited near action buttons,
- the action-details panel now closes correctly when switching between space tabs.
- Playground/log parity are now aligned, so action result wording is consistent across both environments and debugging is easier.
- Location actions on unshared folders now return a clear permission-related error instead of a confusing “ID not found” message.
This Wrike AI agent automatically handles the first setup steps when a new project is created from a blueprint. It assigns unassigned sub-tasks, calculates due dates based on project start date and urgency, and sends a kickoff email to the right CSM, eliminating repetitive manual project setup.
👉 Solution and configuration
Combining This Month’s Features
One of the most practical combinations this month is Approval + Email.
Imagine a creative-review agent set up on a folder of client briefs:
Action 1
Trigger: Task moves to “Ready for Review”
Scope: The triggered item
Agent reads the brief and launches an approval flow, selecting approvers from user chips, for example, the creative lead, plus the legal team if the brief references regulated content.
Action 2
Trigger: Approval is completed
Scope: The triggered item
Agent sends an email through Outlook to the client with the approved brief attached.
The result is a fully automated end-to-end workflow: the review is completed, the client is notified, and no manual handoff is required in between. This is exactly the kind of workflow April’s feature set makes possible.
Questions?
- Need help getting started with any of these features? Reach out in the comments.
- Looking for a different kind of agent? Contact your Customer Success Manager.
- Found a bug or want to share feedback? Use the feedback link in the product or let us know in the comments.
On Our Radar
Here are some of the areas we’re actively exploring next:
- Hierarchy-wide context: Expanding visibility beyond the current 3-level sub-item limit so agents can reason across deeply nested folder structures.
- Data Hub record-change triggers: Enabling agents to trigger when updates from external systems, such as SAP or CRM platforms flow into Wrike via Data Hub records.
- Description change trigger: Supporting agent execution when the task description itself is edited, not just when fields are changed.
- Richer context retrieval: Using graph-based queries to pull related entities into the agent’s reasoning context more effectively, reducing the need for prompt engineering.
Your feedback continues to shape the roadmap, so keep testing, keep sharing what works, and keep telling us what doesn’t
