Spark is the most visible part of Nexthink's AI strategy and the one with the clearest, most measurable impact. It is a fully autonomous AI agent that employees interact with through Microsoft Teams or Microsoft Copilot. When an employee has an IT problem (slow laptop, application not opening, VPN issues, printer won't connect), they message Spark, and Spark diagnoses and fixes the problem without escalating to the service desk.
What makes Spark different from a chatbot or a knowledge base is that it has real-time access to the employee's device telemetry through the Nexthink Collector. It doesn't just suggest generic troubleshooting steps: it reads the actual state of that specific machine: what processes are running, what's consuming CPU, what the network quality is, whether a specific service has failed. The diagnosis is precise because the data is precise.
After diagnosing the problem, Spark executes a fix using Nexthink Remote Actions: IT-approved scripts that run on the endpoint. "IT-approved" is the critical phrase: Spark only executes actions that have been explicitly pre-authorized by IT through the Infinity platform. IT teams define the remediation library Spark can draw from. Spark cannot take actions outside that approved set, which means IT retains full control of what the AI does on managed devices.
The 77% number in context: Nexthink reports that Spark achieves 77% first-contact resolution: meaning 77% of issues employees bring to Spark are fully resolved without any human IT involvement. Nexthink compares this with an industry average of about 15%. That comparison depends on how first-contact resolution is defined: MetricNet’s
2023 service desk benchmark puts the average for human-staffed desks at about 70%. Both figures come from Nexthink’s early-adopter program and launch materials, so validate them in your own pilot. Spark’s real advantage is the real-time device context that a human agent on the phone does not have.
How a Spark interaction works
1
Employee messages Spark
In Microsoft Teams or Copilot, the employee describes the problem in natural language: "My laptop has been really slow since this morning." No ticket required, no hold queue.
2
Spark reads device telemetry
Spark queries the Nexthink Collector on the employee's specific device: checking CPU load, memory usage, running processes, recent crashes, network quality, and application performance in real time.
3
Spark diagnoses the issue
The AI correlates the telemetry against known patterns to identify the most likely root cause. It may ask one clarifying question if the diagnosis is ambiguous, but typically acts directly on what the data shows.
4
Spark executes the fix
If a pre-approved remediation action exists, Spark runs it on the endpoint. The fix takes effect silently in the background: clearing a stuck process, restarting a service, flushing DNS, reconfiguring a setting. Average resolution time: under 2 minutes.
5
Spark confirms and escalates if needed
Spark tells the employee what it found and what it did. If the fix didn't resolve the issue or no approved action covers the problem, Spark hands off to the service desk: with full diagnostic context already documented, reducing the agent's resolution time.
What Spark handles well
Performance issuesHigh CPU/memory, slow startup, fan noise: Spark identifies the offending process and can terminate or restart services.
Connectivity problemsVPN not connecting, Wi-Fi dropping, DNS failures: Spark reads network quality metrics and can flush caches, restart network adapters, or reconnect VPN.
Application failuresApp won't launch, keeps crashing, stuck update: Spark checks process state, recent crash logs, and can repair or restart applications.
Software provisioningMissing application, expired license, software not available: Spark can trigger software installation workflows through integration with UEM platforms.
Password and access issuesAccount lockouts, MFA problems, access requests: handled through integration with identity providers and ITSM workflows.
