Status
Live
Production-ready
Growth
Growth Experiment Agent
Proposes A/B test experiments from uncertain intelligence decisions and recurring issues without clear solutions. Translates low-confidence recommendations into structured experiment proposals with hypotheses, control/va
Intelligence & AnalyticsLiveInternal (Colaberry Enterprise)Verified
Department
Growth
Growth department for Colaberry Enterprise agents
Source
Internal (Colaberry Enterprise)
Built by Colaberry
About
About the Agent
What this agent does, the challenges it addresses, and where it delivers value.
Proposes A/B test experiments from uncertain intelligence decisions and recurring issues without clear solutions. Translates low-confidence recommendations into structured experiment proposals with hypotheses, control/variant definitions, success metrics, and traffic allocation.
Challenges This Agent Addresses
- 1**Marketing**: When the system is unsure whether a campaign config change will help, it proposes a controlled A/B test instead of making an untested change
- 2**Product**: Identifies recurring user experience issues and proposes experiments to test alternative approaches
- 3**Data Science**: Converts uncertain analytical findings into structured experiments with measurable outcomes
Workflow
How the Agent Works
Step-by-step operational flow showing how this agent processes tasks end-to-end.
1
Step 1
Queries the decision table for proposed decisions with moderate confidence (40 to 70)
2
Step 2
For each uncertain decision, generates an experiment proposal with a hypothesis based on the recommended action and target metric
3
Step 3
Searches vector memory for recurring issues that lack clear solutions (similarity threshold above 0.5)
4
Step 4
Creates additional experiment proposals for persistent issues using alternative approaches
5
Step 5
Returns the full list of proposals, limited to 5 from decisions plus memory-based proposals
Execution Modes
Trigger: cron (scheduled periodic analysis)
Data
Inputs & Outputs
What data this agent consumes and the artifacts or actions it produces.
Input Data
- Proposed `IntelligenceDecision` records with confidence scores between 40 and 70
- Vector memory entries flagged as recurring issues without clear solutions
Deliverables
- List of `ExperimentProposal` objects, each containing:
- Hypothesis statement
- Control description (current state)
- Variant description (proposed change)
- Target metric
- Duration in hours
- Traffic split ratio
- Source decision ID (when applicable)
Core Tasks
- Strategic Intelligence
Integrations
Systems Connected
Internal systems, APIs, and tools this agent integrates with.
Tools & APIs
Reads from **IntelligenceDecision** table (proposed decisions)Queries **Vector Memory** for recurring unresolved issuesProposals can be executed by **Execution Agent** (launch_ab_test action)Works alongside **Risk Evaluator Agent** to ensure experiments meet safety thresholds
Specifications
Agent Specs
Technical specifications, requirements, and deployment details.
Related
Related Agents
Other agents in the same department or industry.
Enterprise AI
Ready to deploy this agent?
Schedule a walkthrough with our team to see how this agent integrates with your workflows.