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

# AI Resolution

Investigate recording-linked Zendesk tickets and give agents evidence-backed private notes to review.

AI Resolution investigates an eligible Zendesk ticket after a customer submits a Screendesk recording. It combines the conversation, recording analysis, your workspace guidance, and trusted knowledge to prepare a private note for your support team.

The result can include a supported recommendation and a suggested customer reply, or indicate that human investigation is needed.

{% hint style="info" %}
The current workflow is for **recording-linked Zendesk tickets**. It does not investigate every incoming ticket, send public replies, or automatically solve or close tickets.
{% endhint %}

## Before you enable it

Open **Settings → AI Resolution**. The feature must be available for your workspace, and you need permission to manage its settings.

The **General** tab shows the setup checklist:

* **Zendesk connection:** an active Zendesk integration.
* **Private-note safety:** Screendesk's automatic Zendesk public-reply setting is off.
* **Recording intelligence:** AI processing and report intelligence are available and enabled.
* **Published guidance:** workspace instructions have been published.
* **Trusted knowledge:** at least one trained, usable documentation source.

**Similar solved tickets** is optional. It adds case history but does not replace trusted knowledge.

If the feature is unavailable or shows **Paused by Screendesk**, contact Screendesk support. Changing workspace settings cannot override platform availability.

## Set up the workflow

### 1. Connect Zendesk

Follow [Zendesk installation and setup](/helpdesk-integrations/zendesk/install-and-set-up.md).

In Screendesk's Zendesk integration settings, turn off automatic public replies before enabling AI Resolution. AI Resolution uses private notes for its output.

This setting is separate from a Zendesk trigger that sends the customer a recording-request link.

### 2. Enable recording analysis

In **Settings → AI & Intelligence**, enable **AI processing** and **Report intelligence**. See [AI settings](/debug-with-ai/ai-settings.md).

### 3. Add trusted knowledge

Use [Data Sources](/debug-with-ai/data-sources.md) to connect current troubleshooting documentation, files, or text snippets. Complete training and verify that the relevant content is available.

The current investigation searches **Public** and **Internal** trusted documentation. Restricted sources are excluded. Imported resolved tickets are searched separately as precedent.

### 4. Publish guidance

Open **AI Resolution → Guidance → Add new guidance**.

| Field                     | What to specify                                 |
| ------------------------- | ----------------------------------------------- |
| **Response style**        | Tone and level of detail                        |
| **Required evidence**     | What must be established before proposing a fix |
| **Safety boundaries**     | Claims or actions the agent must avoid          |
| **Escalation rules**      | Cases that need a human investigation           |
| **Privacy and redaction** | Information to omit or redact                   |

For example, require the relevant product version before recommending a version-specific fix, or escalate cases where the recording contradicts the documentation.

Click **Publish version**. New investigations use the active version. Existing attempts retain the guidance they started with, and you can inspect earlier versions in **Version history**.

### 5. Set policy and enable

On **General**, review:

* **Minimum confidence:** below this threshold, the result requires human review. The default is **75%**.
* **Zendesk status tags:** optional tags to help your team track investigation outcomes.
* **Advanced timing:** the evidence deadline and optional transcript wait.

The default evidence deadline is **900 seconds** and optional transcript wait is **300 seconds**. The transcript wait must not exceed the evidence deadline.

Confidence is a decision threshold, not a guarantee that a recommendation is correct. Trusted documentation and the other evidence checks still apply.

Enable AI Resolution and click **Save changes**. Check the setup checklist and status.

## What triggers an investigation?

The workflow starts from a received recording associated with a Zendesk ticket. Supported sources include Zendesk recording requests, Zendesk forms, and Zendesk chat recordings once the ticket association is established.

Screendesk waits for the required recording evidence and analysis. It does not start a ticket-only investigation just because a ticket exists, and a generic recording with no eligible Zendesk association does not qualify.

## How the evidence is used

The investigation considers:

1. The current Zendesk ticket and its comments.
2. The recording evidence and completed report intelligence.
3. Published workspace guidance.
4. Relevant trusted documentation.
5. Similar solved tickets, when available, as lower-trust precedent.

A supported recommendation must cite trusted documentation. Low confidence, conflicting evidence, insufficient documentation, or an escalation requirement can lead to **Needs human**.

## Review the result in Zendesk

The output is a private **AI Resolution** note. Depending on the result, it can include the recording summary, confidence, supporting evidence, and a **Recommended reply**.

Review the note and open the evidence before acting. Any recommended reply is a draft inside the private note; your support agent decides what to edit and send.

If the current ticket changes during the investigation, the older attempt can be marked **Superseded** so stale advice is not published.

## Inspect history in Screendesk

Open **AI Resolution → History** to find investigations by ticket, recording title, or attempt reference. Filter by status and date range, then open an attempt to inspect its outcome and evidence.

| Status                   | What it means                                       |
| ------------------------ | --------------------------------------------------- |
| **Waiting for evidence** | Required recording evidence is still being prepared |
| **Investigating**        | The ticket and evidence are being analyzed          |
| **Recommendation ready** | The evidence checks support a recommendation        |
| **Needs human**          | Human investigation is required                     |
| **Note posted**          | The private note was published to Zendesk           |
| **Superseded**           | A newer ticket state made the attempt stale         |
| **Failed**               | The investigation or publication could not complete |

## If no note appears

Start with the setup checklist, then check the recording's Zendesk association and processing state.

In **History**, inspect the attempt's status and reason. A missing transcript alone is different from missing required report intelligence. An invalid provider key, disconnected Zendesk account, unavailable knowledge, or disabled AI processing can also prevent completion.

AI Resolution adds one investigation note per attempt and checks delivery to avoid duplicate notes on retries. Separate recording receipts or report-intelligence notes may also appear on the ticket.

See [AI in your helpdesk](/debug-with-ai/ai-in-your-helpdesk.md) for the full agent workflow.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.screendesk.io/debug-with-ai/ai-resolution.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.