ControlUp vs Nexthink (2026)

Pulse AI Resolve creates a prepopulated ServiceNow ticket on escalation. Integrations cover ServiceNow and a named technology catalog (Microsoft, Citrix, Omnissa, Amazon, IGEL, Nerdio, Parallels, Intel, thin client vendors), and Workflows has its own integrations list in the API. Device actions and scripts run on the customer's endpoints and cloud hosts through ControlUp's own agents.

Sourcecontrolup.com/platform/ai-powered-it-self-serviceread 2026-09-08

Remediations run on the employee's own device, and Spark makes an authenticated round trip into the customer's ITSM. Its actions, remote actions, agent actions, workflows and others, run during a conversation through the Nexthink Collector, with their results fed into the next turn, so the changes land in the customer's estate, not in Nexthink's own records. ServiceNow is named for knowledge-base and ticket indexing and for the service request catalog.

Spark retrieves past incidents from a third-party ITSM system and raises escalated tickets back into it for the employee, with a short description, the actions taken and the conversation transcript, and it can retrieve and link to third-party data when the customer asks. Underneath sit a native connector system, published APIs and the Nexthink Query Language, and Amplify brings DEX context and remediations straight into the ITSM console. Spark is also embedded in tools such as MS Teams and ServiceNow Virtual Agent.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card and nexthink.com/platform/integrations-infinityread 2026-09-12

The AI Assistant acts as an orchestrator, interpreting intent and carrying out complex multi step actions across platform, synthetic monitoring and workflow areas. The platform has a named workflow runtime, Workflows, with flows, runs, forms and integrations in the API and a low code automation workflow builder on every plan. Pulse AI runs guided or autonomous remediation through it, in a loop that detects, decides and resolves.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Spark picks its own tools in a loop until it needs the employee. Its LLM can select one or more tools to gather information, querying device and user data, connecting to local data sources through the Collector, searching knowledge articles, the service request catalog and the available remote actions, agent actions and workflows, and retrieving past incidents from ITSM.

It then offers or executes actions, escalates or ends the conversation, repeating the cycle until it needs input from the employee, so tools are chosen at run time, not along a coded branch. Nexthink calls Spark an agentic AI resolution engine that understands employee issues in natural language, evaluates DEX data and system context, and executes IT-approved actions to resolve issues on its own.

Flow adds low-code, event-driven workflows that customers write to detect, communicate, integrate and act, and remote actions, agent actions and workflows are configuration objects the customer creates and enables. Workspace serves as a central agentic cockpit for IT.

Each participant has a distinct role, from the employee who asks and the administrator who sets the approved actions to the service desk agent who receives an escalated ticket with the transcript and actions taken, as in an example that runs from a slow-laptop complaint through diagnosis to remediation or escalation.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card, /platform/user-guide/spark and nexthink.com/platform/flowread 2026-09-12

The API exposes triggers, trigger schedules and trigger scripts for VDI and DaaS, device alerts and synthetic monitoring alert policies. Pulse AI Resolve alerts end users on their desktop when thresholds are breached, and Pulse AI Insights surfaces anomalies. Work reaches the assistant in the console, through the on device Connect client and through ServiceNow escalation.

Sourceapi.controlup.io/referenceread 2026-09-08

Employees reach Spark in Teams and other enterprise chats, including MS Teams, ServiceNow Virtual Agent, Moveworks and Amazon Connect, with further reach through APIs and the A2A protocol into whatever chat front end the customer already runs. Nexthink ships its own Microsoft Teams application and works in LLM-supported languages, with English or Japanese as the tenant fallback.

Separately, Flow runs event-driven workflows fired by real-time alerts from endpoint telemetry, not by an employee starting a conversation, and alerting supports proactive IT management. Work therefore arrives two ways, from an employee asking in a channel and from a condition detected on a device. Outbound messaging and surveys reach employees in-app, and Spark sends progress updates and questions during a resolution, not only a final answer.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card and nexthink.com/platform/flowread 2026-09-12

Pulse AI grounds answers on the customer's live and historical environment data, queried across modules over MCP, on ControlUp's own public knowledge base, and on files a user attaches to a conversation. The persistent structure it queries is operational telemetry held in the platform's data indices. There is no index over the customer's own documents, runbooks or knowledge base, and ControlUp does not say whether the Self Service Hub can draw on the customer's own knowledge sources.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Spark retrieves from an index of the customer's own knowledge base, past tickets and resolution notes, alongside live device telemetry. Its sources include knowledge base articles the customer imports into Nexthink Infinity, past ticket insights, and human-authored resolution notes from third-party systems that help Spark reduce escalations. ServiceNow knowledge base articles and tickets are indexed for retrieval, and manually uploaded files are indexed too, so the index stands between runs.

On each run, Spark's LLM picks the tools it needs to search knowledge articles and the service request catalog and to pull past incidents from a third-party ITSM system, suggesting remediations from notes written by human agents. As an optional fallback that an administrator must enable, Spark can search a curated list of trusted websites chosen by Nexthink when customer-specific sources fall short. Users can see the sources behind a response to check its accuracy.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-cardread 2026-09-12

Follow up prompts on the same topic carry context within an AI Assistant conversation. Conversations are not currently saved, and conversation history is still being built. Pulse AI learns normal patterns, naming conventions and the best remediation steps, and that learning is built into the product, not kept as a memory layer.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Working notes last only for the conversation, and no memory carries over between conversations. Spark writes internal notes, from the user, to keep context through multi-step interactions, and in A2A integrations context is carried by reusing contextId and taskId values within a conversation thread. Device timelines and endpoint state are the customer's operational data, held because monitoring devices is the product.

Spark's improvement over time comes from supervisor feedback used in reinforcement learning, which is learning folded into the model. The 30 day conversation previews and 13 month summaries are audit records kept for supervisors. Nexthink's AI model card lists every input, output, step and storage period Spark uses and names no persistent agent memory, preference store or decision trace across conversations.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card and /platform/configuring_nexthink/.../spark-agent2agent-integrationread 2026-09-12

The AI Assistant asks the user for approval before running any action. Actions need both the Allow AI Assistant Actions permission and the permission for the specific action, the assistant can only do what the user could already do in the product UI, and administrators can turn it off for a user or role. Pulse AI remediation runs as guided workflows or autonomously by policy, within guardrails IT defines.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Before an impactful fix runs, the employee sees its impact and has to answer yes, and this consent step is on by default. Spark runs only within IT-defined guardrails using approved actions, and for actions an administrator flags as needing employee consent, such as remediations with an impact, Spark shows the impact from the action configuration and runs the action only after an explicit yes. Spark always asks for user approval before taking issue-resolution actions unless an admin has explicitly allowed actions to run without consent.

Customers vet and select which remote or agent actions Spark can use, and an action is available only if it was enabled during Spark configuration. Spark escalates a ticket to the support team only after the employee confirms, and supervisors review interactions for quality and give feedback. Inference passes through AWS Bedrock Guardrails, which filter six harm categories and block prompt attacks, and Spark cannot answer about other users or their devices.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card and /platform/user-guide/sparkread 2026-09-12

ControlUp holds SOC 2, audited by EY with the type not stated, a public SOC 3 report, CSA STAR Level 1, and ISO 27001, 27017, 27018 and 27701 with certificates linked. It uses FIPS 140-2 validated modules, FedRAMP authorization is in progress, and an AI security white paper sits on the trust center. Access runs through SAML sign on and SSO groups, role based access control with custom roles, an IP allowlist and an audit log, all exposed in the public API, and AI queries and actions run under the user's exact permissions.

Sourcecontrolup.com/controlling-your-securityread 2026-09-08

Admins use role-based access control to give specific users or groups access to Spark's AI features while restricting others, and employee access is controlled through the admin console of the channel Spark is embedded in. Secure access tokens are used, and an employee can reach only their own devices and data. Data moves over HTTPS and is encrypted with AES-256 at rest, personal data fields are removed automatically when NQL queries are generated, retention is configurable by the customer, and a DPA is available.

The trust center shows a certification strip with the AICPA mark and four more certification images labeled only Certification 1 through 4, so those standards cannot be named. The CEO's signed statement names compliance with GDPR, CCPA and the EU AI Act, and a dedicated AI Compliance Team of legal, privacy and security experts reviews each AI component. The Security Hub at security-hub.nexthink.com is where the four unlabeled certifications would be named.

Sourcenexthink.com/trust-center and docs.nexthink.com/legal/global-ai-hub/spark-ai-model-cardread 2026-09-12

An audit log of the platform is exposed through the public API, workflow runs have a queryable status, and every AI Assistant response shows the data sources and analysis used. ControlUp's FAQ says in one answer that query history sits in audit logs for 90 days, and in another that queries are not currently saved in audit logs, with conversation history still being built. The agent's reasoning cannot yet be reconstructed after the fact.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Audit logs cover Spark's conversations, decisions and remediations, and admins define approved actions, workflows and governance policies and review those logs. Spark can log its internal reasoning and processes for supervisor review without sharing them with employees, and reasoning details for supervisors are a standard system output. Escalated tickets carry the actions taken and the conversation transcript to the service desk.

Full conversation previews, including employee PII, are kept for 30 days and conversation summaries with a user reference for 13 months, under retention settings the customer configures. Service desk supervisors see Spark-employee conversations and resolution-performance metrics, and employees can check the sources used to build a response. Nexthink's real time observability across devices, networks and applications covers the customer's estate and is a separate product from Spark's own logs.

Sourcedocs.nexthink.com/platform/user-guide/spark and /legal/global-ai-hub/spark-ai-model-cardread 2026-09-12

AI data is processed in the same region as the customer's ControlUp deployment, with US and EU named, and ControlUp says it never crosses the geographic boundaries set in the service agreement. Agents installed on the customer's endpoints are the edge of the same deployment. How the region is chosen is not stated.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Customer data, including AI processing, stays in the customer's region. All data is stored in the customer's region, Spark uses GenAI models that Nexthink operates in its own AWS environment in that region, and AI processing on AWS takes place in the AWS region aligned with the customer's Nexthink deployment. No region list, named customer environment or way to choose a region is published. Deployment is cloud only on AWS, the Collector is an endpoint agent that feeds that cloud, not a deployment option, and there is no self-hosted or on-premise option.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card and nexthink.com/trust-centerread 2026-09-12

A public Script Library is a browsable catalog of ready to run remediation and diagnostic scripts that a customer can adopt as they are, linked from every page's Tools menu and listed by the Scripts endpoint in the API. Remediation actions and scripts are included on every plan, and Pulse AI ships its own remediations for common issues.

Sourcecontrolup.com/scriptsread 2026-09-08

Nexthink's automation catalog of remote actions, agent actions, workflows and others is one of Spark's knowledge sources, and customers vet and select which of them Spark can use, since actions are available only if enabled during Spark configuration. Each remediation is a unit a customer switches on and Spark uses when it fits, and removing one leaves the rest working.

A customer-facing Library is published at nexthink.com/library and promoted in the site navigation, though its size and whether entries come from Nexthink or the community are not described. Flow is a low-code engine for building further workflows, and the platform is sold as Infinity plus named expansion modules and named AI agents.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card and nexthink.com/libraryread 2026-09-12

The AI Assistant uses a variety of models from Azure AI Foundry, optimized for different types of queries and analysis and hosted in Microsoft's enterprise environment, and ControlUp routes between them. Customers and admins cannot choose a model.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Spark uses vetted, fine-tuned off-the-shelf LLMs, including Meta Code Llama and Claude by Anthropic, and runs agentic reasoning through AWS Bedrock LLM APIs, with Bedrock as the hosting and inference layer, not a separate model. Nexthink picks the model for each component, and there is no model picker, admin setting or bring your own key option. Nexthink does not train models across customers, builds no global training set from customer interactions, and shares no customer or personal data with the AI providers themselves, such as Anthropic or Meta.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-cardread 2026-09-12

A public API reference at api.controlup.io covers organizations, users, roles, SAML, SSO groups, audit log, events and alerts, device actions and scripts, VDI triggers with schedules, workflow flows and runs, synthetic monitoring and DaaS IQ. Customers generate keys that inherit the user's permissions, and PowerShell cmdlets are available. The ControlUp MCP server lets a customer's own licensed agents invoke ControlUp product tools and operations.

Sourceapi.controlup.io/referenceread 2026-09-08

Through the Spark Agent2Agent integration, outside agents can discover and invoke Spark over A2A, an open standard, to query device and user insights, run safe guardrailed actions, receive asynchronous callbacks and present clear outcomes. Spark exposes its agent card to a primary AI agent, which routes the employee request to Spark as a subagent, with contextId and taskId carried across turns and responses returned to a configured webhook over authenticated service-to-service communication.

A step-by-step setup guide covers this public vendor-neutral API. The platform documentation at docs.nexthink.com includes a legal and AI hub and an llms.txt index at docs.nexthink.com/legal/llms.txt with markdown versions of each page, alongside the Nexthink Query Language, published APIs and Amplify, which brings Nexthink context into the customer's ITSM. Spark also runs inside outside systems such as MS Teams and Moveworks, which call Nexthink.

Sourcedocs.nexthink.com A2A integration guide and /legal/global-ai-hub/spark-ai-model-cardread 2026-09-12

Customers have no way to evaluate a change to the AI Assistant, a workflow or a Pulse AI remediation. ControlUp monitors the assistant's accuracy and response quality itself. Synthetic monitoring tests measure the customer's applications, DaaS IQ's validate VM settings check validates a deployment configuration, and App Profiler and ROI tooling measure the customer's estate, not the agents.

Sourcesupport.controlup.com/v1/docs/ai-assistantread 2026-09-08

Nexthink validates Spark model updates against test datasets, using precision, recall and proprietary in-house metrics suited to each part of the system, so a change is tested and the results can be compared. A post-deployment loop runs on top for each customer. Supervisors assess Spark conversation logs and give feedback that feeds reinforcement learning cycles, and that learning is used only for the customer's own employees, not pooled.

The system tracks resolution success rate and escalation frequency, and supervisors get resolution-performance metrics. DEX scoring, benchmarking against fifteen million endpoints and AI adoption measurement look at the customer's estate, not the agent. The precision-and-recall harness is Nexthink's own and gates its model updates, and no sandbox where a customer can test a changed action or workflow before it reaches employees is described.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-cardread 2026-09-12

Agents act on endpoints through ControlUp's own installed agents, scripts and API actions, and the AI Assistant operates the ControlUp platform itself. Live Remote Management and Remote Assist let a human administrator control an endpoint through the console. No agent operates a browser, desktop or computer.

Sourcecontrolup.com/pricingread 2026-09-08

Devices are fixed through scripted actions run by an endpoint agent, not by operating a screen. Spark runs remote actions, agent actions, workflows and others through the Nexthink Collector on the device and connects to local data sources through the Collector for troubleshooting.

This is machine-to-machine execution against processes, services, configuration and files, not windows and buttons, so a changed screen layout does not affect it. No browser session, desktop control or RPA fallback appears among Spark's inputs, outputs and data flow.

Spark can search a curated set of external web resources when customer knowledge falls short, which is retrieval from a list Nexthink curates.

Sourcedocs.nexthink.com/legal/global-ai-hub/spark-ai-model-card and /platform/user-guide/sparkread 2026-09-12