> For the complete documentation index, see [llms.txt](https://docs.wrangle.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.wrangle.io/reporting/ticket-reporting/ai-deflection-reports.md).

# AI Deflection reports

<figure><img src="/files/b7cDVOMctyPzEaRJmF9T" alt=""><figcaption></figcaption></figure>

### Filters

The following filters are available at the top of the page for deflection reports:

* **Date Range** — preset shortcuts or a custom from/to range
* **Inbox** — filter to one or more inboxes
* **Tags** — filter to tickets with specific tags (depends on selected inboxes)
* **Priority** — filter by ticket priority

### Charts view

The default view shows a summary of outcomes based on users' answers to the prompt below, shown after every deflection attempt made by Wrangle:

<figure><img src="/files/QH2KJ6kmo2xgwjKgOdxa" alt="" width="350"><figcaption></figcaption></figure>

#### Outcomes pie chart

<figure><img src="/files/YE7mwjNjLDV4ipLZZrX8" alt="" width="563"><figcaption></figcaption></figure>

The chart — **"Outcomes After AI Ticket Deflection Attempt"** — breaks every deflection attempt into one of four outcomes:

<table><thead><tr><th width="282.9765625">Outcome</th><th>What it means</th></tr></thead><tbody><tr><td><strong>No Response from User</strong></td><td>The user was shown an AI response and did <strong>not</strong> click either either button.</td></tr><tr><td><strong>Confirmed Helpful</strong></td><td>The user clicked "👍 <strong>Yes, this helped.</strong>"</td></tr><tr><td><strong>Rejected Response</strong></td><td>The user was shown a response but clicked "👎 <strong>No, start a ticket.</strong>" to create a ticket.</td></tr><tr><td><strong>No KB or Smart Reply Match</strong></td><td>No matching content was found, so Wrangle AI did not post a message; a ticket was created automatically</td></tr></tbody></table>

#### Summary cards

<figure><img src="/files/pSLd02esAOZDSl9ysmuW" alt="" width="375"><figcaption></figcaption></figure>

To the right of the chart, four metric cards provide totals for the selected period:

* **Tickets Confirmed Deflected** — count of "Confirmed Helpful" outcomes
* **Total Ticket Deflections** — "Confirmed Helpful" plus "No Response from User" (deflections where the user may not have confirmed but did not create a ticket)
* **Tickets Created: After Response** — tickets created by users who saw a response but chose to create a ticket anyway
* **Tickets Created: No Match** — tickets created automatically because no matching knowledge base content or Smart Reply was found

{% hint style="info" %}
For **After Response** and **No Match**, a downward trend is a positive signal — it means fewer tickets are being created despite deflection attempts.
{% endhint %}

***

### Table view

Switch to **Table View** using the toggle above the chart. The table shows one row per deflection attempt — every question your AI responded to — paginated in sets of 25. This is where you can dig into the details behind your summary numbers.

<figure><img src="/files/sU4hYq85YeDJJwA7yipS" alt=""><figcaption></figcaption></figure>

#### Result

Each row shows a color-coded **Result** that tells you what happened after the AI responded:

* **Confirmed** (purple) — the user clicked to confirm the response answered their question
* **Response needed** (dark purple) — the user didn't reply and didn't create a ticket; the outcome is unclear
* **Created ticket** (orange) — the user saw the response but submitted a ticket anyway
* **No match** (gray) — no knowledge base content or Smart Reply matched, so a ticket was created automatically

If a ticket was created, the Result column becomes a link — click it to jump directly to that ticket.

#### Question and Response

The **Question** column shows what the user actually asked, and the **Response** column shows exactly what the AI sent back. Long text is truncated by default; click "more" to expand inline. Reviewing these side by side is the fastest way to spot where your knowledge base has gaps or where responses aren't landing well.

#### Sources

The **Sources** column shows which knowledge base articles or Smart Reply keywords were used to generate the response. Up to two sources are shown per row, with an option to expand if there are more. For AI Ticket Deflection responses, each source is a clickable link to the original article. For Smart Replies, you'll see the keyword that triggered the reply.

#### Type

The **Type** column distinguishes between **AI Ticket Deflection** (AI-generated responses from your knowledge base) and **Smart Replies** (keyword-triggered responses). Viewing by type helps you evaluate each system's performance separately.

#### Rating

{% hint style="info" %}
The **Rating** column lets you score the quality of any AI response on a 1–5 star scale, directly from the table. Click the stars on any row to submit a rating.
{% endhint %}

Ratings are a lightweight way for your team to flag responses that are misleading, off-topic, or just not quite right — without having to open a ticket or leave the report. Over time, consistently low-rated responses point to specific knowledge base articles that may need to be updated or removed. This is one of the most actionable signals in the report for improving deflection quality.

***

### Exporting data

The **Export AI Usage Data** button downloads a CSV of both AI Ticket Deflection and Smart Reply usage for the selected date range and filters.