> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fieldpal.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Setting up AI report review

> How to tell FieldPal what a good report looks like, so it can check each report before it is submitted or grade it afterwards.

AI report review checks a filled-in report against standards you write for its template. You describe what a good answer looks like, in the same words you would use to brief a new starter, and FieldPal reads the technician's answers and photos against that description.

You choose when the review runs:

| Option | When it runs | Who sees the result |
| - | - | - |
| **Before submit** | When the technician taps **Review** in the app | The technician, who can fix their answers or submit anyway. You see it on the report afterwards. |
| **After submit** | In the background, once the report is submitted | Admins only. The app gives the technician no sign that the report is graded. |
| **Off** | Never | Nobody. This is the default. |

To see what the technician sees in the app, read [AI report review](https://docs.fieldpal.ai/app/ai-review.md) in the app guide.

> [!NOTE]
>
> Review settings are part of the new template editor, which FieldPal turns on for each organisation. If your template editor doesn't have a **Review** section, ask your FieldPal contact.

## Open the review settings

1. In the left-hand navigation, click **Report templates** and open the template.
2. In the list of sections down the left of the editor, click **Review**. Once checks are set up, it shows how many there are. The section is headed **What a good report looks like**.
3. Under **When to review**, choose **Before submit**, **After submit** or **Off**.

![Review section of the Warranty Claim template, with Before submit selected and the first two of four checks listed](https://mintcdn.com/fieldpal/5JP2Ckxot4lw2rUK/images/ai-review/dashboard-01-review-section.png?fit=max\&auto=format\&n=5JP2Ckxot4lw2rUK\&q=85\&s=49f41f73e528b6206d2f9c33402a17e4)

Each check in the list shows its name, its description, which fields it **Looks at**, and a **Must pass** tick box you can change without opening the check.

## Add your checks

Each check is one thing that gets a report sent back today: a vague fault description, a missing reading, a blurred photo. Click **+ Add check** and fill in:

* **Name**: a short label, such as *Cause and diagnosis*. With **Before submit**, the technician sees this name in the app.
* **What a good answer looks like**: a plain description of a good answer. FieldPal checks the answers and photos against this. The technician never sees it.
* **Looks at**: tick the fields this check judges. Tick nothing and the check looks at the whole report. Photo fields you tick are sent to FieldPal as images, so a check can judge what a photo shows.
* **Weight**: how much the check counts towards the score, from 1 to 10.
* **Must pass**: tick this for anything essential. If a must-pass check fails, the whole review fails, whatever the score. In the app, these checks are marked **Required**.

![The Cause and diagnosis check open for editing, showing Name, What a good answer looks like, the Looks at field list with Cause ticked, Weight 2 and Must pass ticked](https://mintcdn.com/fieldpal/5JP2Ckxot4lw2rUK/images/ai-review/dashboard-02-check-settings.png?fit=max\&auto=format\&n=5JP2Ckxot4lw2rUK\&q=85\&s=68a0dffdc08ee47f28f2b5b69a12fe12)

Click **Edit** on a check to change it, and **Done** to close it. Use **Up** and **Down** to change the order, and **Remove** to delete one. A template can have up to 20 checks.

> [!TIP]
>
> Write one check per problem rather than one long check for everything. Each check gets its own row in the result, so the technician can see exactly which answer to fix.

## Try the checks on a real report

Under **Try these checks on a real report**, click **Pick a report**, choose a report filled in from this template under **Report to try**, and click **Run review**. FieldPal judges that report against your checks as they stand, including changes you haven't saved yet. It shows the headline with the score, the summary, and whether each check was **Met** or **Not met**, with a fix for any that weren't.

![Try these checks on a real report, with a submitted Warranty Claim picked and the result Checks passed at 100%, every check marked Met](https://mintcdn.com/fieldpal/5JP2Ckxot4lw2rUK/images/ai-review/dashboard-04-try-result.png?fit=max\&auto=format\&n=5JP2Ckxot4lw2rUK\&q=85\&s=2759de395794ea904805b76e7409a578)

Trying a check saves nothing, either to the template or to the report. The template has to be saved, with at least one report filled in from it, before you can try it. You can run up to 10 tries a minute.

## Scoring

**Scoring**, **What crews see** and **Tone and strictness** sit below the checks. Each one shows a short summary of its current setting, such as *Pass mark 70 · score hidden from crews*. Click one to open it.

![Scoring, What crews see and Tone and strictness opened, showing a 70% pass mark, the score hidden from crews, the default headlines and an empty guidance box](https://mintcdn.com/fieldpal/5JP2Ckxot4lw2rUK/images/ai-review/dashboard-03-scoring-and-wording.png?fit=max\&auto=format\&n=5JP2Ckxot4lw2rUK\&q=85\&s=67761af49246e69b66a2e73d56a07579)

Open **Scoring** to set:

* **Pass mark**: the score a report needs, from 0% to 100%. The default is 70%. The score is the weighted share of checks the report meets. A report passes when it reaches the pass mark and meets every must-pass check.
* **Show the score to crews**: with **Before submit**, whether the technician sees the score in the app. Admins always see it. With **After submit**, technicians never see a score.

## What technicians see

With **Before submit**, open **What crews see** to change the headline at the top of the result in the app:

* **Headline when a check fails**: the default is *Some checks didn't pass*.
* **Message when it passes**: the default is *Checks passed*.

Leave either blank to use the default, which shows greyed out in the box.

## Tone and strictness

Open **Tone and strictness** to add optional **Guidance for the review**: how strict to be and how to word advice. For example, *Be strict about blurred photos. Ignore spelling.*

## Save the template

Click **Save** at the bottom right of the editor. It reads **Saved** when there is nothing new to save. If a check is missing a name or a description, or looks at a field that is no longer on the template, the editor shows **Can't save yet** with the problem. Click **Fix in Review** to go straight to it.

> [!WARNING]
>
> A report keeps the checks its template had when the report was started. Changing or removing checks affects reports started afterwards, not reports already in progress or submitted. A report started before its template had any checks is never reviewed.

## Where results appear

* **On the report**: every reviewed report shows an **AI review** panel in the dashboard. See [viewing reports](https://docs.fieldpal.ai/dashboard/viewing-reports.md#ai-review-results).
* **In the report email**: when a report passed a **Before submit** check, or was submitted after one failed, the submitted-report email repeats the check, with the checks that weren't met listed first.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.