# 03 — Now Assist Document Intelligence

> **Release:** Zurich | **Flow:** Requestor Flow — Phase 1 (Steps 4)

***

## What It Is

**Now Assist in Document Intelligence (NADI)** uses generative AI to extract structured data from images and documents and map it directly into ServiceNow table fields — with no manual data entry required.

In Zurich, the legacy Document Intelligence (DocIntel) application is **no longer activated on new instances** and is being prepared for future deprecation. **Now Assist in Document Intelligence is the current, supported product** for all new implementations.

In this lab, NADI is configured with a use case called **Veritas Extract**. When a user uploads an error screenshot or device label image (via the Conversation Topic upload in Step 4 of the Requestor Flow), NADI auto-triggers on the Incident attachment and extracts structured fields — specifically `error_code` — which arms the downstream Agentic Workflow trigger.

***

## Role in the Requestor Flow

![NADI Flow Overview](/files/ynKIEq8KmJbkh9jDuqEo)

***

## What NADI Enables in This Lab

| Capability              | How NADI Delivers It                                                               |
| ----------------------- | ---------------------------------------------------------------------------------- |
| Auto-field population   | Error code and device details extracted from uploaded image — no manual copy-paste |
| Agentic Workflow arming | `error_code` populated on the Incident, enabling downstream resolution             |
| Full automation mode    | No agent review required — GenAI writes directly to record fields                  |
| Richer AI agent context | Extracted error code used by Resolution Pathfinder to search KB, logs, and web     |
| Higher data quality     | AI reads directly from source image — eliminates transcription errors              |

***

## Lab Exercise — Steps to Configure NADI

### Step 1: Open Now Assist in Document Intelligence

Navigate to **All** → search **Now Assist Admin** → **Skills** → **Platform** → **Search for 'Document'** → **Edit 'Extract information from documents'**

![NADI — Now Assist Features screen](/files/Iio52cFB1orp8thq1vPB)

> This is the entry point for all NADI configuration. The skill is OOB — you do not create it. You create **use cases** within it.

***

### Step 2: Create the Use Case

1. Click **Edit** on the **Extract Information from documents** skill
2. Click **New use case**

![NADI — Create Use Case](/files/Wz139VOhrOZl5JMMKPIr)

| Field             | Value                                                        |
| ----------------- | ------------------------------------------------------------ |
| **Use case name** | `Veritas Extract`                                            |
| **Target table**  | `Incident Extend table (x_snc_apacaienable_incident_extend)` |
| **LLM**           | `Azure OpenAI - GPT Large` (or your configured provider)     |
| **Image Mode**    | Yes                                                          |

3. Click **Next**

![NADI — Use Case Details](/files/DJEyCQ4l17ku2P8p8eUr)

***

### Step 3: Add Extraction Fields

On the **Fields** step, click **Add a field** — select **Field** (not question or table).

![NADI — Add Fields](/files/FEBHsJ6aOu0t1RPN5wbt)

#### Field 1 — Error Code

![NADI — Error Code Field](/files/hW6SiQcciXuxGLXrsxS1)

| Setting                     | Value        |
| --------------------------- | ------------ |
| **Field name**              | `Error Code` |
| **Field type**              | `Text`       |
| **Target field**            | `error_code` |
| **Required for extraction** | Yes          |

**Details**

```
This is the error code mentioned in the image, example of error code text is "0xE00052", however, we only want to extract 52 from this.
```

> **This is the critical field.** `error_code` gates the downstream Agentic Workflow as it is the most reliable indicator for resolution troubleshooting.

#### Field 2 — Model Details

![NADI — Model Details Field](/files/xNV94wDv8sPU3lPkuA4N)

| Setting                     | Value           |
| --------------------------- | --------------- |
| **Field name**              | `Model Details` |
| **Field type**              | `Text`          |
| **Target field**            | `model details` |
| **Required for extraction** | Yes             |

**Details**

```
This is Reg. Model Version, Alpha numeric field on the image
```

#### Field 3 — Product Name

![NADI — Product Name Field](/files/4V4ly91RPR6mMgIjKdD3)

| Setting                     | Value          |
| --------------------------- | -------------- |
| **Field name**              | `product name` |
| **Field type**              | `Text`         |
| **Target field**            | `product name` |
| **Required for extraction** | Yes.           |

**Details**

```
This is the product name which is mentioned on the device details image, generally mentioned below "Product:"
```

#### Field 4 — Serial Number

![NADI — Serial Number Field](/files/2DK5km14cvFrLJGwawoX)

| Setting                     | Value           |
| --------------------------- | --------------- |
| **Field name**              | `Serial Number` |
| **Field type**              | `Text`          |
| **Target field**            | `serial number` |
| **Required for extraction** | Yes             |

**Details**

```
This is the SN number mentioned at the end of the image on the left side, mentioned after the text "SN:"
```

#### Field 5 — PN Bar Code

![NADI — PN Bar Code Field](/files/sz6DeIMfPKO2zlhpPGcB)

| Setting                     | Value              |
| --------------------------- | ------------------ |
| **Field name**              | `PN Bar Code`      |
| **Field type**              | `Text`             |
| **Target field**            | `product bar code` |
| **Required for extraction** | Yes.               |

**Details**

```
This is the pn number above the bar code, which is alpha numeric also it is hyphenated
```

***

### Step 4: Test the Extraction

This step validates that NADI can correctly read your lab images and extract the configured fields before activating for production use.

#### Step 4a — Upload the Test Images

In the **Test output** step, an upload dialog appears.

![NADI — Test upload dialog](/files/cFdYb8A2M4mxoxKY5GLt)

Select **Upload from this device** and click **+ Add file**.

> **Add both lab images:**
>
> * The **Error 37 screenshot** (error screen from the Veritas appliance)
> * The **Veritas server details label** (device label with model, serial, product details)
>
> Both images are required to test all 5 extraction fields. If you missed downloading the lab resources, they are available in the Prerequisites section.

Click **Continue**.

#### Step 4b — Review Extracted Fields

NADI processes both images and opens the **Document Q\&A** panel.

![NADI — Extracted fields pending review](/files/0TPmaZWZWIwNFRUOZ7wn)

The panel shows:

* **All (5)** tab — all 5 configured fields
* **To review (5)** tab — fields pending your review
* Each field shows its extracted value below the field name
* Fields not yet reviewed appear with a grey spinner icon

Example extracted values: `error code` = `37`, `model detais` = `VER5000W`

#### Step 4c — Mark All Fields as Reviewed

For each field, click **Mark As Reviewed** to confirm the extracted value is correct.

![NADI — Mark As Reviewed button](/files/BJyiOeArKtRG6MWaw4cs)

> Mark all 5 extracted fields as reviewed. Each field card turns green and the **To review** count decrements.

#### Step 4d — Verify All Fields are Green

![NADI — All fields reviewed and green](/files/ZGHuUDRePHbfWfMBqyny)

* **All fields should be green** — each field card shows a green checkmark
* **To review count should become 0** — confirmed you have reviewed all fields

> If any field shows an unexpected value, click the edit (pencil) icon to correct it before submitting.

#### Step 4e — Confirm and Submit

Click **Submit** (top right). The **Confirm field predictions** dialog appears.

![NADI — Confirm and submit dialog](/files/ddJFTUoDlyDiH88b4OhS)

The dialog reads: *"Confirm that you've reviewed all fields for accuracy before you submit."*

Click **Confirm and submit**.

> The error code field (`37` in this example) is the key value — this populates `u_extracted_error_code` on the Incident and triggers the Resolution Pathfinder Agentic Workflow.

***

### Step 5: Add the Integration

1. Navigate to the **Integrations** step in the use case
2. Click **Add Integration**

![NADI — Integration config 1](/files/CYGT23AhKjfxgLBgzZ24)

| Field                | Value             |
| -------------------- | ----------------- |
| **Integration Name** | `Veritas_Process` |
| **Target Table**     | `incident extend` |
| **Integration type** | `Process task`    |
| **Create Flow**      | Checked           |

**Condition** — configure when the flow should fire:

![NADI — Process integration condition](/files/5r6voV4gHac6fCPFhWrv)

| Field    | Operator       |
| -------- | -------------- |
| `Number` | `is not empty` |

> This ensures the integration only fires on records with a valid incident number.

3. Ensure **Create Flow** is checked, then click **Save**

![NADI — Integration config 2](/files/zb5AqrKTZ5eRR2sByCew)

This auto-generates:

```
Flow: DocIntel Extract Values Flow — Veritas Extract
Trigger: Document Task Updated WHERE status = Done
```

***

### Step 6: Activate the Integration

1. Locate **Veritas\_Process** in the Integrations panel

![NADI — Activate integration](/files/xZbik7NzQrLUEwLaKFpV)

2. Click **Open in Flow Designer**
3. Click **Activate** — status updates to **Active**

***

### Step 7: Configure the Incident Integration Trigger

1. Navigate to **Integrations** tab of the use case
2. Add a second integration:

![NADI — Integration trigger config 1](/files/9UfsrZFZkln0mADrmwz4)

| Field                | Value             |
| -------------------- | ----------------- |
| **Integration Name** | `Veritas_Extract` |
| **Target Table**     | `incident extend` |
| **Integration type** | `Extract Values`  |
| **Create Flow**      | Checked           |

![NADI — Integration trigger config 2](/files/0oJrpr9W4gQDOJLrvb48)

3. Verify the trigger activates correctly

![NADI — Activate trigger step 1](/files/OQ4V0XdOOy4ziG9JLNav)

![NADI — Activate trigger step 2](/files/4U1U6P3DjgvHsYiYgiCx)

***

### Step 8: Review and Activate the Use Case

Review the full configuration on the **Review and activate** screen (Step 5 of the wizard).

![NADI — Review and activate summary](/files/rXa2UtiopIKdhfb6kagb)

**Verify use case details:**

| Field         | Expected value    |
| ------------- | ----------------- |
| Use case name | `Veritas Extract` |
| Target table  | `incident extend` |

**Verify all 5 fields are present:**

| Field name    | Type | Required |
| ------------- | ---- | -------- |
| error code    | Text | Yes      |
| model details | Text | Yes      |
| product name  | Text | Yes      |
| serial number | Text | Yes      |
| pn bar code   | Text | Yes      |

**Verify both integrations are present:**

| Integration name               | Type           |
| ------------------------------ | -------------- |
| `Veritas_Extract Integrations` | Extract Values |
| `Veritas_Process`              | Process Task   |

Click **Complete setup**.

![NADI — Use case set successfully](/files/4hOenTqbXz0xOJwzEn24)

> **"Your use case has been set"** — Click **Return to use cases**.

***

### Step 9: Verify Full Automation End-to-End

1. Click the **Settings** (gear) icon on the use case
2. Navigate to **Extraction mode**
3. Toggle **Full automation mode (no agent review required)** to **On**

![NADI — Full automation verification](/files/nFLj2uhHV7hBqXZoJeIK)

![NADI — Full automation result](/files/UgsMpspg4suyPrGENP18)

> Full automation mode means GenAI extracts and writes fields immediately — no agent review required. This is required for the Requestor Flow so the Agentic Workflow trigger can fire within the same session.

***

## Key Configuration Fields

| Field             | Value for This Lab                 |
| ----------------- | ---------------------------------- |
| Skill             | Extract Information from documents |
| Use case name     | `Veritas Extract`                  |
| Target table      | Incident extend table              |
| Extraction mode   | Full automation (no agent review)  |
| Integration 1     | `Veritas_Extract` — Extract Values |
| Integration 2     | `Veritas_Process` — Process Task   |
| Process condition | Number is not empty                |

***

## Technical Notes

### Full Automation Mode vs. Agent Review Mode

| Mode                | Behaviour                                        | Use When                                    |
| ------------------- | ------------------------------------------------ | ------------------------------------------- |
| **Full automation** | GenAI extracts and writes immediately, no review | Consistent document layouts (device labels) |
| **Agent review**    | Human reviews before values are written          | Complex or variable documents               |

***

## Reference

* [ServiceNow Zurich — Now Assist in Document Intelligence](https://www.servicenow.com/docs/bundle/zurich-intelligent-experiences/page/administer/document-intelligence/concept/docintel-nowassist-landing.html)
* [Use cases for Now Assist in Document Intelligence](https://www.servicenow.com/docs/r/intelligent-experiences/now-assist-in-document-intelligence/use-cases-now-assist-document-intelligence.html)

***

## Next Step

Continue to [04a — Now Assist Skill Kit: CreateOptimalSearchQuery](/sn-ai-lab/phase-2-fulfiller-flow/04a-now-assist-skill-kit-createoptimalsearchqueryskill.md) to build the first of three Custom Now Assist Skills needed for the Resolution Pathfinder Agent.


---

# Agent Instructions: Querying This Documentation

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

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://sn-ai-platform.gitbook.io/sn-ai-lab/phase-1-requestor-flow/03-now-assist-document-intelligence.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
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.