> ## Documentation Index > Fetch the complete documentation index at: https://docs.tierzero.ai/llms.txt > Use this file to discover all available pages before exploring further. # Support Agent > AI-powered internal support that investigates live systems to answer engineering questions ## What is a Support Agent? Get unblocked instantly. Stay in flow forever. TierZero Support Agent responds to queries in your #ask-eng and #ask-infra channels. It goes beyond searching docs to investigate live systems, cross-reference telemetry, and store tribal knowledge from past answers. Tribal knowledge scales without burning out the people who hold it. ## How It Works Engineers ask questions directly in Slack. "How do I roll back service X?" "What's the runbook for database failover?" No context-switching, no ticket filing, no waiting. TierZero searches Notion, Confluence, runbooks, code repos, past incidents, and Slack history. It goes beyond docs by cross-referencing live telemetry, deployment state, and code to give answers that are actually current. Track deflection rate, CSAT, and question categories out of the box. See which topics need better documentation, which answers get thumbs-down, and fine-tune the agent's behavior without writing code. ## Key Capabilities ### Investigates Live Systems Not just doc retrieval. Support Agent queries your actual infrastructure: * Build logs and CI/CD pipelines * Observability platforms (Datadog, New Relic, etc.) * Recent code commits and deployments * Service configurations and dependencies ### Per-Channel SOPs Maximum customizability per team. Configure different behaviors for different channels: * \#ask-eng can focus on code and deployment questions * \#ask-infra can prioritize infrastructure and system health * Each channel can have its own instructions and knowledge sources ### Insights Dashboard See what's working and tune what isn't: * Question analytics and trending topics * Deflection rate tracking * Knowledge gap detection, highlighting topics where the agent struggles * CSAT scores from team feedback ### Standard Operating Procedures Execute common ops tasks directly from Slack with guardrails: * Restart pods, clear caches, scale deployments * Junior engineers handle routine operations safely * Senior engineers stop being a human API ### Gets Smarter Over Time Learns from feedback: * Thumbs up/down on responses improves future answers * Common questions get faster, more accurate responses * New tribal knowledge is captured automatically ## Use Cases ### Reduce Interruptions for Senior Engineers * **Challenge**: Senior engineers spend hours answering the same questions repeatedly. * **How it works**: Support Agent monitors your ask channels and responds with accurate, context-rich answers by investigating live systems. * **Outcome**: Senior engineers stay focused on building while the team still gets expert-level answers. 85% of questions answered without escalation. ### Faster Onboarding * **Challenge**: New team members don't know where to find information or who to ask. * **How it works**: New engineers ask questions in Slack and get immediate, detailed answers with links to relevant code, docs, and systems. * **Outcome**: Faster ramp-up time and reduced dependency on specific team members. Median response time under 30 seconds. ### Preserve Tribal Knowledge * **Challenge**: Critical knowledge lives in people's heads and is lost when they leave. * **How it works**: Every question and answer becomes part of TierZero's knowledge base, building institutional memory over time. * **Outcome**: Knowledge persists even as team composition changes. ## How to Create a Support Agent ### Step 1: Select Slack Channel Choose the Slack channel where your support agent will respond to messages. * Go to Support Agents and click "Create Support Agent" * Select a public Slack channel to monitor * The channel must be public and TierZero must be able to join it ### Step 2: Review Instructions & Categories Review and customize the auto-generated instructions and categories for your support agent. * TierZero analyzes the channel history to understand the types of questions asked * **Instructions**: Review and edit the auto-generated instructions that guide how the agent responds * **Categories**: Review the auto-detected question categories for your channel * Wait for processing to complete before continuing ### Step 3: Review Review your configuration and create the support agent. * Set a name for your support agent * Review all settings * Click "Create Agent" ## After Creation ### Monitoring Responses * View all conversations in the Support Agent details page * See response quality metrics and feedback * Track deflection rate and CSAT scores ### Refining Instructions * Edit instructions based on response quality * Add or modify categories as new question types emerge * Use feedback to continuously improve ## Best Practices **1. Start with High-Volume Channels** * Begin with channels that get the most questions * This maximizes impact and provides more data for improvement **2. Review Auto-Generated Instructions** * The auto-generated instructions are a starting point * Customize them based on your team's specific needs and terminology **3. Encourage Feedback** * Ask team members to react to responses (thumbs up/down) * Use feedback to continuously improve answer quality **4. Monitor Early Responses** * Pay close attention to the first few days of responses * Make adjustments quickly based on any issues you see ## Example Instructions [Support Agent Instructions](/prompt-library/support-agent-instructions)