Research Intelligence Evaluator Agent

Research Intelligence Evaluator Agent is a reference agent blueprint for teams dealing with research and strategy teams need synthesis across large source sets with explicit provenance, tradeoffs, and update tracking. It is designed to score outputs against explicit rubrics so teams can compare variants, regressions, and rollout quality over time.

Where It Fits

• Domain: Research Intelligence
• Core stakeholders: research teams, strategy leads, executives
• Primary tools: document corpus, search index, source tracker

Operating Model

1. Intake the current request, case, or workflow state.
2. Apply evaluation logic to the available evidence and system context.
3. Produce an explicit output artifact such as a summary, decision, routing action, or next-step plan.
4. Hand off to a human, a downstream tool, or another specialist when confidence or permissions require it.

What Good Looks Like

• Keeps outputs grounded in the most relevant internal context.
• Leaves a clear trace of why the recommendation or action was taken.
• Supports escalation instead of hiding uncertainty.

Implementation Notes

Use this agent when the team needs briefing memos, source comparison, trend monitoring with tighter consistency and lower manual overhead. A good production setup usually combines structured inputs, bounded tool access, and a review path for high-risk decisions.

Suggested Metrics

• Throughput for research intelligence workflows
• Escalation rate to human operators
• Quality score from evaluation review
• Time saved per completed workflow