> For the complete documentation index, see [llms.txt](https://docs.klink.cloud/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.klink.cloud/dashboard-and-reports/analytics-dashboard/ai-dashboard.md). # AI Dashboard {% hint style="info" %} This page explains the **AI** tab on the Analytics Dashboard, especially **AI Resolution Rate**, how it should be calculated, and how it differs from Chat Dashboard Resolution Rate. {% endhint %} ## Where to find it 1. Open **Analytics Dashboard**. 2. Switch to the **AI** tab. 3. Pick a **date range**. 4. Export charts with **⋮**, then **Export chart** (PNG or CSV) when needed. The AI tab typically covers: | Metric | What it answers | | ---------------------- | --------------------------------------------------------------------------- | | **AI Resolution Rate** | What share of AI-handled conversations did the AI close on its own? | | **AI-assisted rate** | How often did AI help (suggest or reply) even if a human finished the case? | | **Response speed** | How fast did AI first respond? | | **Deflection savings** | Rough value of conversations handled without a human | | **Revenue impact** | Optional commercial impact views when configured | *** ## Screen tour ### KPI overview ![AI Dashboard KPI overview](https://420146782-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FqqwHJh6YA4uvzfFRLnoR%2Fuploads%2FGMUFDLZaGhdp8yy6lQJO%2Fkpi-overview.png?alt=media\&token=8fb53f87-2cb0-411a-9406-c867c505c5af) At the top of the AI tab you typically see: * **KPI cards:** AI Cases Resolved, AI Resolution Rate, AI-Assisted Rate, Human-Only Resolution Rate, and Median AI Response. * **AI vs Human** donut (share of cases resolved by AI vs humans). * **AI Cases by Channel** (for example Chat vs other channels). * **Daily AI vs Human** stacked bars for case volume over the selected range. Use these cards to sanity-check volume and mix before reading the trend charts below. ### AI Resolution Rate trend ![AI Resolution Rate Trend chart](https://420146782-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FqqwHJh6YA4uvzfFRLnoR%2Fuploads%2FRbxVmuhL0uCtYUR94Ppq%2Frate-trend.png?alt=media\&token=4bb85ef3-e5b9-4b32-ac2d-4ed37a37a9df) This chart usually combines: * A **line** for AI Resolution Rate (often a **7-day rolling** average). * **Volume bars** under the line so you can see how many AI-related cases each day contributed to that day's percentage. A sharp drop with stable product usage often points to a definition or filter issue (workflows, disposition mapping) rather than a real drop in AI close success. Check recent workflow and disposition changes first. ### Impact charts ![AI impact charts](https://420146782-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FqqwHJh6YA4uvzfFRLnoR%2Fuploads%2FBcdCfK9DKNSI7AejEQcp%2Fimpact-charts.png?alt=media\&token=5d941815-fbfe-4b56-a399-d57c08b9f7f8) Lower on the page you typically find: * **Case Volume AI vs Human** (daily or period comparison). * **AI Impact on Service Level** (how AI-handled volume relates to SLA or service-level views). * **AI Resolution by Hour** (when in the day AI closes cases). * **AI Resolution by Day of Week** (weekday patterns). These views help spot staffing, channel, and time-of-day patterns alongside the headline resolution rate. *** ## AI Resolution Rate ### What it is **AI Resolution Rate** is the percentage of conversations that were **resolved by the AI agent** (Kai / Chat AI Agent), without needing a human agent to take over and close the case. Many workspaces show this as a **7-day rolling** line (daily rate smoothed over the last 7 days), with daily case volume as bars. ### How it should be calculated **Intended definition (product intent):** `AI Resolution Rate = (Cases closed by AI / AI-handled cases in the period) × 100%` **Closed by AI** means the case was auto-closed by the AI lifecycle, for example **AI Agent Session Timeout** (or equivalent AI auto-close / AI resolved status). It is **not** whether a human wrap-up form later shows a custom disposition. **AI-handled cases** means conversations that entered the AI agent path in that period (bot was active on the room). {% hint style="warning" %} **Important: dispositions and workflows** AI Resolution Rate must **ignore wrap-up disposition values** set by workflows. If you add a workflow that changes disposition to a **custom disposition** when the AI times out or auto-closes a room, the AI Resolution Rate chart should **not** drop. Those rooms are still AI-resolved. If your rate falls right after enabling such a workflow, the metric is likely filtering on disposition incorrectly. Contact support or your CSM and ask engineering to recalculate using AI close status (for example AI Agent Session Timeout) rather than disposition. {% endhint %} ### What it is not * Not the same as **Chat Dashboard → Resolution Rate** (that page describes human or agent chat resolution). * Not CSAT, sentiment, or SLA Meet rate. * Not "every closed case with any disposition named Resolved." ### Reading the trend chart * **Line:** rolling AI Resolution % (often 7-day). * **Bars:** daily case volume for the AI-related population. Taller bars mean that day's % is based on more cases. * A sudden drop with **stable product usage** usually means a **definition or filter bug** (workflows, disposition mapping) or a real drop in AI close success. Check recent workflow and disposition changes first. ### Historical data after a fix When engineering corrects the metric: 1. Ask whether the fix is **forward-only** or includes a **backfill / recalculation** from the date the bad filter started. 2. Until a backfill runs, numbers in the broken window may remain wrong even after the code fix. 3. Pre-change periods (before the disposition workflow) are usually still trustworthy. *** ## Related dashboards | Page | Use when | | ---------------- | ----------------------------------------------------- | | Chat Dashboard | Human or agent chat volumes, AFRT, ART, member status | | Call Dashboard | Voice queue and abandon metrics | | Ticket Dashboard | Ticket volumes and statuses | | CX Log | Case-level history, dispositions, CSAT, sentiment | *** ## Troubleshooting | Symptom | Likely cause | What to do | | ----------------------------------------------- | ------------------------------------------------------------------------------------------------- | ------------------------------------------------------- | | AI Resolution Rate drops after a new workflow | Workflow sets custom disposition on AI auto-close; metric wrongly requires a specific disposition | Confirm with eng; metric should key off AI close status | | Rate looks fine but CX Log shows many AI closes | Dashboard filter does not match CX Log filters (channel, queue, date TZ) | Align date range, timezone, and queue filters | | Rate jumps on low-volume days | Small denominator (few AI cases) | Prefer 7-day rolling; check volume bars | Need help interpreting a specific tenant chart? Reach out to or your CSM.