When your team runs the same type of analysis repeatedly (competitive landscapes, asset deep dives, cross-trial comparisons), Workflows let you codify the methodology, expected deliverables, and quality standards into a reusable template. Think of them as SOPs for the AI agent: you define how the work should be done, and the agent executes it consistently every time.
Maven includes built-in workflows for common analysis types:
Search & Screen | Filtered entity lists with prioritization criteria |
Deal Intelligence | Transaction analysis with comparable deal terms |
NPV Analysis | Probability-weighted revenue models with sensitivity tables |
TPP Development | Target Product Profile comparison tables with min/target values |
Cross-Trial Comparison | Indirect comparison tables with efficacy/safety benchmarks |
Opportunity Assessment | Market sizing with TAM/patient population estimates |
Landscape Analysis | Competitive positioning narratives with strategic implications |
Asset Deep Dive | Single-drug profiles covering MOA, clinical data, and commercial outlook |
Company Profile | Pipeline, BD activity, strategy, and financials overview |
Indication Overview | Disease primers with pathophysiology and treatment algorithms |
Meeting Prep | Counterparty briefings with recent events and strategic priorities |
Access workflows
Click Workflows in the left toolbar, or type / in the task input field to see available workflows in a dropdown. Workflows are organized into three categories:
My Workflows: Personal workflows visible only to you
Organization Workflows: Shared workflows available to your team
Maven Library: Pre-built workflows from Maven Bio
Create a new workflow
Click Create Workflow to begin a four-step creation process.
Step 1: Workflow Details. Enter the workflow name, set visibility (Private or Organization-wide), and write a trigger description. The trigger description tells the agent when to suggest or activate this workflow. A good trigger description is specific: "Use when the user asks for a competitive landscape in an oncology indication with 5+ active programs."
Step 2: Deliverables. Define what the workflow should produce. Select a type from the dropdown (Smart Table, Research Report, Excel Model, Financial Analysis, Graphics, PPTx, PDF, or Free Text), give it a name, and add a description of what the output should contain. You can optionally upload a template (PPT format, structured document, financial model, or any reusable organizational format) to standardize the output. For each deliverable, configure whether it is a Primary Output, Required Output, both, or neither.
Step 3: Methodology. Outline the step-by-step process the agent should follow. Include the research steps, data sources to consult, analytical structure, required calculations or frameworks, and any logic flow. This becomes the execution blueprint.
Step 4: Quality Criteria. Define validation requirements the agent must satisfy before completing the workflow: completeness checks, formatting requirements, data integrity checks, and structural expectations.
Save and publish
After completing all steps, you can Save as Draft to keep the workflow editable and inactive, or Publish to make it available for use. Toggle the workflow status to Published when ready.
Use a workflow
From the agent input box, type your query, then type / and select a workflow from the dropdown. The agent executes the selected workflow according to the methodology and deliverables you defined. You can also click one of the quick-access workflows displayed below the input box.
What you can do next
Start by running one of Maven's built-in workflows to see the output format, then customize it or create your own. To learn how to delegate tasks to the agent (including attaching files and using inline tagging), see Delegate your Tasks to the AI Agent. To share workflows with your team, see Managing and Sharing Permissions for Blueprints.