> For the complete documentation index, see [llms.txt](https://docs.nexthink.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nexthink.com/user-guide/spark/working-with-spark.md).

# Working with Spark

Spark is a conversational IT agent that serves as the first point of contact for employees when they need IT assistance.

Unlike traditional chatbots that rely primarily on predefined scripts or static knowledge bases, Spark follows an agentic approach. It can reason over employee context and execute IT-approved actions to diagnose and remediate issues.

Spark leverages extensive device context and Digital Employee Experience (DEX) data, enabling more accurate analysis and targeted resolution. By combining contextual awareness, action capabilities, and learning from resolved incidents, Spark accelerates issue resolution and reduces the need for manual IT intervention.

## Spark architecture

**Spark** connects with employee requests through the communication channel, runs a diagnosis, and attempts issue resolution. The image shows a system architecture diagram that illustrates how Spark, an AI-powered support system, connects various components and handles employee requests end-to-end.

<figure><img src="https://268444917-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FxJSUDk9NTtCHYPG5EWs3%2Fuploads%2F9EjiwIjiPGJRUMMLBbVQ%2FSpark_architecture_20261105161003.png?alt=media&amp;token=b1fac666-0a27-4f11-a1ad-43c30f066fe2" alt=""><figcaption></figcaption></figure>

## Context and data inputs

To provide relevant responses, Spark uses a combination of static and dynamic data sources.

### **Contextual Nexthink data and capabilities**

Spark has access to the data available through the standard NQL data model. This includes device health, diagnostics, user metadata, remediations, Adopt guides, and newly released capabilities. Spark also has access to most of the dynamic data model, such as custom fields and data generated by remote action executions. Spark only accesses data relevant to the employee it is interacting with.

Spark relies on [specific NQL data model tables](https://docs.nexthink.com/user-guide/spark/spark-nql-capabilities.md) to query Spark-user interaction data.

### **Knowledge base articles**

Spark has access to knowledge base articles [manually imported](https://docs.nexthink.com/user-guide/administration/ai-resources/managing-knowledge-sources.md) from the ITSM or ingested via the [ServiceNow Knowledge Base connector](https://docs.nexthink.com/configuring_nexthink/bringing-data-into-your-nexthink-instance/integrating-nexthink-with-third-party-tools/inbound-connectors/connector-for-servicenow-knowledge-base.md).

### **Service request catalog**

Spark uses catalog structure, forms, and request metadata ingested from ServiceNow to identify and recommend service requests and provide associated guidelines.

Refer to [ServiceNow Request Catalog connector](https://docs.nexthink.com/configuring_nexthink/bringing-data-into-your-nexthink-instance/integrating-nexthink-with-third-party-tools/inbound-connectors/connector-for-servicenow-request-catalog.md) for more information.

### **Live device diagnostics**

Real-time data from the user's device. Spark runs live device diagnostics through the Nexthink Collector when it needs the device's current state to troubleshoot the issue. Spark mentions only the results that are relevant to the issue, and Nexthink does not store the full output.

Refer to [Live Device Diagnostics](https://docs.nexthink.com/understanding-key-data-platform-concepts/live-device-diagnostics.md) for more information.

### **Past ticket resolution data**

Spark use resolution notes for incidents resolved by support agents to suggest remediations and continuously reduce manual interventions and escalations.

Refer to [ServiceNow Tickets connector](https://docs.nexthink.com/configuring_nexthink/bringing-data-into-your-nexthink-instance/integrating-nexthink-with-third-party-tools/inbound-connectors/servicenow-tickets-connector.md) for more information.

### **Curated list of trusted websites**

If enabled, Spark can search a curated list of trusted web domains when internal knowledge sources do not provide sufficient guidance. This capability helps Spark resolve a broader range of issues autonomously.

Refer to [Managing Spark settings and data inputs](https://docs.nexthink.com/user-guide/spark/setting-up-and-managing-spark/managing-spark-settings.md#enabling-web-search-in-spark) for more information.

{% hint style="info" %}
Personal data handling is covered under the Nexthink Data Processing Agreement (DPA). Spark processing is user-specific and restricted to the customer region.

Spark never provides data from other organizations.
{% endhint %}

## Adaptive learning

Spark learns on its own, with no configuration, and no fine-tuning. Each day, Spark improves from the issues it resolved, so your day-180 agent is measurably sharper than your day-1 agent.

{% hint style="info" %}
Adaptive learning is enabled by default. There is nothing to configure, enable, or manage—it works as soon as Spark starts resolving issues.
{% endhint %}

Adaptive learning in Spark guarantees:

* **Faster resolutions for employees:** Recurring issues resolve in fewer back-and-forth turns as Spark learns what works.
* **Higher first-contact resolution:** Recurring issues resolve on the first attempt more often as Spark learns from experience.
* **Consistency across employees:** Spark doesn't start from scratch each time it troubleshoots a problem it handled before with another employee.
* **No ongoing work:** The system improves automatically as Spark does its job. The more you use Spark, the better it gets.

### The autonomous loop <a href="#user-content-the-autonomous-loop" id="user-content-the-autonomous-loop"></a>

Spark adaptive learning is a simple, self-reinforcing cycle:

1. **Spark resolves**: Every conversation produces a verified trail: what the problem was, how Spark diagnosed it, what action it took, and whether it worked.
2. **Adaptive Learning reflects**: Spark consolidates those resolved cases into learned patterns, capturing what worked, what didn't, and what pitfalls to avoid.
3. **Handled from experience**: The next time a similar issue arrives, Spark draws on what it learned. It recognizes the pattern faster, asks fewer questions, and applies the proven fix.
4. **Resolution compounds**: Each resolved case makes Spark sharper for the next one. The more Spark handles, the more it learns.

## Spark Actions

Spark can perform actions to diagnose and remediate employee issues. These actions are enabled by Nexthink administrators after IT approval. Before executing a remediation action, Spark requests confirmation from the employee. Built-in diagnostic actions are excluded from this requirement and may run in the background without interrupting employee work.

Refer to [Managing Spark actions](https://docs.nexthink.com/user-guide/spark/setting-up-and-managing-spark/managing-agent-actions.md) for more information.

***

RELATED PAGES:

* [Understanding employee experience with Spark](https://docs.nexthink.com/user-guide/spark/working-with-spark/understanding-employee-experience-with-spark.md)
* [Monitoring Spark](https://docs.nexthink.com/user-guide/spark/working-with-spark/monitoring-spark.md)
* [Setting up and managing Spark](https://docs.nexthink.com/user-guide/spark/setting-up-and-managing-spark.md)
* [Using Recommendations to enhance Spark knowledge and actions](https://docs.nexthink.com/user-guide/spark/working-with-spark/using-recommendations-to-enhance-spark-knowledge-and-actions.md)


---

# 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.nexthink.com/user-guide/spark/working-with-spark.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.
