customer-research

The customer-research skill acts as a comprehensive research assistant for support agents and customer success managers who need to synthesize information across disparate internal and external systems. Designed to facilitate efficient problem-solving, this skill allows users to investigate complex technical issues, retrieve account-specific context, or gather background information necessary for drafting high-quality customer communications. By systematically querying connected data sources—ranging from official documentation and CRM notes to team communication channels and live web data—it provides a structured, authoritative answer while maintaining strict transparency regarding the reliability of the findings.

• Performs multi-tier systematic research: prioritized searching across official internal sources (knowledge bases, product docs), organizational records (CRM notes, support tickets), team collaboration history (Slack, email, calendars), and external web resources.

• Synthesizes findings into a standardized, reader-friendly brief including direct answers, evidence-backed supporting points, and clear confidence scoring (High/Medium/Low).

• Highlights critical context, nuance, and potential caveats such as roadmap availability, security constraints, or conflicting information found across different tiers.

• Identifies research gaps, suggesting follow-up actions or subject matter experts when connected sources yield insufficient data.

• Facilitates immediate next-step actions including drafting customer responses, suggesting knowledge base updates, or creating new runbook entries for institutional learning.

• Operates by parsing the specific nature of a request (question, investigation, or context search) and prioritizing authoritative documentation over speculative team communication.

• Users should provide specific inputs like a customer question, reported bug ID, or account name to trigger the search workflow.

• Be aware that confidence levels are directly tied to the source tier: internal documentation (Tier 1) receives the highest trust, whereas inferences and analogies (Tier 5) are clearly flagged as low-confidence and require human verification.

• When results are inconclusive, the agent is trained to be transparent, explicitly asking the user for missing context or suggesting internal experts rather than hallucinating details.

• Useful for preparing for customer meetings, troubleshooting recurring issues, or maintaining consistency in support history by reviewing previous communications across the account lifecycle.