
Ep232: From On-Prem to AI-Native: Nexthink’s Big Bet on Autonomous IT
Shownotes
Nexthink’s journey from on-prem to cloud to AI-native shows how two decades of endpoint data, Amazon Bedrock and rigorous evaluations power Spark, an autonomous agent that is reshaping the service desk.
Topics Include:
- Nexthink: two decades turning workplace IT failures into fixable signals.
- Monitoring 25 million endpoints, processing gigabytes of real-time signal per second.
- AI-native revenue jumped from $20M to $109M in one year.
- Rewriting every line of code to go from on-prem to cloud.
- Why cloud transformation is about culture and mindset, not just technology.
- Remediation and workflow engines running billions of fixes at scale.
- How ChatGPT's arrival sparked an AI lab to reinvent the company.
- Six months of prototyping that became an autonomous IT agent.
- Every enterprise shares one lasting problem: technology will eventually fail.
- Employees suffer first, work around issues, and rarely report them.
- Giving IT visibility and tools to fix problems before tickets pile up.
- Why data and context, not frontier models, are the real moat.
- "Teams is slow": what generic AI guesses versus what telemetry knows.
- Why everyone hated chatbots: they routed tickets instead of solving problems.
- Spark's success metric: resolve in minutes, escalate only when necessary.
- Grounded in device data, Spark knows more than you tell it.
- Eyes on the problem, hands to fix it, with permissions respected.
- A hotel-room VPN failure diagnosed from a single screenshot.
- Inside troubleshooting mode: hypotheses, diagnostics, and an audited fix in seconds.
- Why Amazon Bedrock enabled a model-agnostic, multi-model strategy.
- When to stick with foundation models and when to fine-tune.
- The harness: where most agent engineering time actually goes.
- Evaluations from day zero: never ship changes on intuition.
- "Let the LLM reason, let the data decide."
- Treating AWS as an engineering partner, not a supplier.
- Scaling Amazon MSK from 5 to roughly 15 gigabytes per second.
- Will the service desk as we know it survive five years?
- Customers reporting 70% resolution rates and 40% ticket deflection.
- A broken-monitor fix that uncovered hundreds of employees with the same unreported problem.
- An email distribution list request handled by Spark in minutes.
Participants:
- Moe Haidar: Head of Agentic AI Engineering (Spark AI), Nexthink
- Kamil Davidov: Sales Leader, Israel ISV-BizApps, Amazon Web Services
See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon.com/isv/