TL;DR: AI agents are now generally available for all Business+ accounts, so there’s no need to access them in Wrike Labs anymore. The biggest new capability is multi-action agents, which let you combine multiple actions in a single agent without chaining. Agents can now also look up records in linked databases. Featured example: A “revision tracker” agent that counts and documents review cycles. It can be deployed in about five minutes and replaces what previously required a recipe in an integration platform.
Hey Community! 👋
Welcome to Wrike Agent News!
This monthly digest features updates, learnings, and how-tos from the Wrike R&D team.
This month is a milestone: AI agents are now generally available 💯 Everything you’ve been testing in Labs is production-ready and available on Business+ plans. Thank you! And the feature you requested most, multi-action agents, is also here.
🚀 Generally available:
AI agents are out of Labs.
Agents are now enabled by default on all Business+ accounts. Admins can manage the feature in Settings → Generative AI.
What this means:
• No more Labs opt-in required.
• Agents are production-ready and supported.
• Admin control via account settings toggle.
• Available on Business, Enterprise, and Pinnacle plans.
If you were already running agents in Labs, nothing changes; your existing agents keep working. If you haven’t started yet, now’s a great time.
🎯 Multi-action agents
One agent. Multiple actions. No more chaining for independent steps.
Agents can now execute two or more actions from a single trigger. Add a prefix, an assignment, and a comment all inside one agent.
Use case: New task created → agent classifies the request, renames it with a prefix like “[DESIGN]”, assigns the right specialist, and posts a summary comment. One agent replaces what previously required three chained agents.
Use case: Status changes to “Signed Off” → agent assigns the next owner and updates a custom field with the sign-off date. Two actions, zero handoffs.
Each action gets its own prompt, plus a new General Instruction field that is prepended to every action. Use the general instruction for shared context: what the agent’s role is, what it should look at, and reference data like team rosters or category lists. Then, each action prompt focuses on its specific job: classify, rename, or assign. Write the context once, and every action has it.
Note: Multi-action means independent actions running in parallel. If Action B needs the result of Action A, you still need agent chaining. Multi-action is for actions that don’t depend on each other.
🔗 Link-to-database action
Agents can now look up and select database records.
Two database upgrades:
- Read: Agents can see Link to Database fields and Mirror fields when analyzing a task, including data pulled from linked records, not just the task’s own fields.
- Write: When a task mentions a client name, project code, or equipment ID, the agent can search your linked database and select the matching record. This automatically populates all associated mirror fields.
Use case: New support ticket mentions “Acme Corp” → agent finds the matching record in your client database, selects it, and mirror fields auto-fill with account manager, SLA tier, and contract details.
Use case: Task description contains equipment ID → agent matches it in the equipment database and populates warranty status, location, and vendor info.
🛠️ Under the hood
Smarter context, better reliability.
- Improved retrieval: Agents now receive cleaner, more relevant context. Irrelevant data is filtered out before the agent reasons, leading to more accurate results.
- Empty fields visible: Agents can now see empty custom fields, so they can correctly identify what’s missing and needs filling.
- Language respect: Agents respond in the same language as their instructions, write your prompt in German, get results in German.
- Peak load handling: Backend improvements ensure agents perform reliably even during high-volume periods.
Agent of the month: Revision tracker
In this section, we’ll dive into a specific use case.
Use case: Teams that collaborate with clients or external stakeholders, agencies, professional services, legal review teams often send work back and forth multiple times before it’s finalized. Tracking how many revision cycles each item goes through is valuable for identifying bottlenecks, managing client expectations, and scoping future work.
“We needed to know how many times a campaign was returned to the client for more information. Before, this would have required a recipe in our integration platform. Now, it’s a three-sentence agent.”
The solution: A single multi-action agent with two actions:
- Action 1 (Update field): Increments a numeric custom field (“Revision Count”) each time the status returns to a review stage. If the field is empty, it sets it to 1. If it already has a value, it adds 1.
- Action 2 (Comment): Posts a note recording the revision, including which cycle this is and a brief summary of what the agent observed in the task at the time of the status change.
Time saved: Eliminates manual tracking and replaces what previously required a recipe in an integration platform.
Prerequisites: A numeric custom field (e.g., “Revision Count”), a workflow with a client-facing review status.
Deployment time: 5 minutes: the prompt is just a few sentences.
How it works:
- Trigger fires: Status changes to the review stage (e.g., “Client Update Needed”).
- Counts: Agent reads the current revision count and increments it by one.
- Documents: Agent posts a comment noting the revision cycle number and context.
Two actions. Automatic revision history. No integration platform required.
Click the link below to view the solution and step-by-step configuration.
👉 Solution and configuration
The complete guide includes:
- General instructions and action prompts with customization instructions.
- Step-by-step deployment walkthrough.
- Examples for agency/client review, legal redlining, and creative approval workflows.
- Tips for adapting the counter pattern to other use cases (e.g., escalation counting, reassignment tracking).
Combining this month’s features
Multi-action agents and agent chaining complement each other. Use multi-action for independent steps, chaining for dependent ones.
Complete task lifecycle: Status changes to review stage → Revision tracker counts the cycle and documents context. If a task gets stuck, the stale task escalation chain from December catches it.
Combine a multi-action intake agent (to triage new work) with the revision tracker (to monitor ongoing cycles) and the escalation chain (to catch stalled items), three agents covering the full lifecycle.
Questions?
Want help setting up the revision tracker? Reach out to us in the comments.
Need a different agent type? Contact your Customer Success Manager.
Found a bug or have feedback? Use the feedback link in the product or let us know in the comments.
On our radar
Areas we’re actively exploring:
- Sequential action dependencies: Actions where Step B uses the result of Step A.
- More trigger types: Expanding what can kick off an agent.
- Expanded field support: Broader coverage of field types that agents can read and write.
Your feedback shapes our roadmap. Keep testing, keep sharing what works (and what doesn’t).
