> ## Documentation Index > Fetch the complete documentation index at: https://docs.getthread.com/llms.txt > Use this file to discover all available pages before exploring further. # Triage Agent Summary Dashboard > Track Triage Agent performance end to end — zero-touch and handoff rates, the triage flow Sankey, outcome trends by month and board, and technician and customer feedback. The Triage Agent Summary dashboard gives you a complete view of how Triage Agent is performing across your service desk — from the moment a ticket arrives to how it ultimately resolves. It answers three questions in one place: how much the AI is actually doing, how well it's doing it, and what people think of the results. Open it from **Thread admin → General → Analytics**. ## What this dashboard answers * What share of tickets is Triage Agent resolving **with no technician involvement**? * Where do tickets **drop out** of the triage flow, and into what? * Is zero-touch performance **improving month over month**? * Which **boards** are getting the most value from triage, and which need configuration work? * What are **technicians and customers** saying about the AI's work? ## Filters and time windows A filter bar at the top applies globally — every chart, table, and metric on the page updates together: | Filter | Options | | ---------- | ---------------------------------------------- | | Date range | Defaults to the last 90 days | | Board | A single service queue, or all boards together | ## What's on the dashboard ### Session overview Headline metrics for the selected period, each carrying a status badge — on target, below target, or high — so a problem is visible without reading the numbers: * **Total sessions** and **unique tickets** triaged * **Zero-touch rate** — resolved by the AI with no technician involvement * **Assisted handoff rate** — triaged by the AI, then routed to a technician * **Intervention rate** — a technician had to step in or reply directly * **Blocked rate** — sessions that failed to complete, through timeouts, errors, or rejected approvals * **AI first reply** — median time to the AI's first response * **Average reply time** and **feedback count** ### Triage flow A Sankey diagram showing how tickets move through triage, with four views you can switch between: | View | What it shows | | ---------- | ----------------------------------------------------------------------------------------------------- | | Population | Every ticket, split into those that got a triage session and those that didn't | | Buckets | Triage sessions broken into their outcome buckets — resolved by AI, handed off, intervention, blocked | | Outcomes | How each bucket ultimately closes — zero-touch, closed, or still open | | No session | What happens to tickets that skip triage entirely | Bar width represents ticket volume, so the bulk of your traffic and the places it drops off are both visible at a glance. ### Bucket breakdown The same outcome buckets as a table — session counts and each bucket's share of total volume, with proportional bars for quick comparison. ### Outcome trend A 12-month line chart tracking zero-touch, closed, and open outcomes month over month. This is where you see whether AI performance is genuinely improving or just holding. ### Board performance Outcome rates — zero-touch, closed, open — broken out by service board, so you can tell which queues are benefiting and which need a configuration adjustment. ### Feedback feed A single chronological feed of every rating on the AI's work, from both audiences: * **Technicians** — routing quality ratings * **Customers** — reactions to AI-generated notes Filter the feed by source (Customers, Technicians) or by type (Routing Ratings). A summary bar shows total ratings, positive rate, and average routing score. ## Metric definitions | Metric | Definition | | ---------------- | -------------------------------------------------------------------------------- | | Session | One Triage Agent run on a ticket. A ticket can have more than one. | | Zero-touch | The ticket resolved without a technician being involved at all. | | Assisted handoff | Triage Agent gathered what was needed and routed the ticket to a technician. | | Intervention | A technician stepped in or replied directly rather than letting triage complete. | | Blocked | The session failed to complete — a timeout, an error, or a rejected approval. | | AI first reply | Median elapsed time from ticket arrival to the AI's first response. | ## Things to watch * **Sessions and unique tickets differ.** A ticket can be triaged more than once, so read rates against the denominator the tile names. * **A high intervention rate isn't automatically bad** early in a rollout — technicians stepping in is how you find the intent gaps. It's a problem when it stays high. * **Blocked sessions are a configuration signal, not a quality signal.** Look at approvals and integration setup before you look at the AI. * **The 90-day default hides a new rollout's progress.** Narrow the date range when you've just changed intents or settings. * **Data refreshes daily**, so today's sessions generally appear tomorrow. ## Next steps Adjust the configuration behind these outcomes. Find the next requests worth automating. ## Related topics - [Voice AI Summary Dashboard](/analytics/voice-ai-dashboard.md) - [Getting Started with Magic Analytics](/analytics/getting-started-with-magic-analytics.md) - [ROI Value Realization Dashboard](/analytics/roi-value-realization-dashboard.md)