A 2024
Gartner survey found that 73% of enterprises experienced at least one AI-related security incident in the past year, with an average cost of $4.8 million per breach. At the same time, “shadow AI,” the unsanctioned use of public AI tools by employees, is creating new risks of data leakage, privacy violations, and compliance failures.
The solution isn’t just locking tools down; it’s giving employees the stack they actually want to use. When IT sanctions modern, visual, AI-assisted apps and invests in onboarding, workers are less likely to go around official channels. Just as important, prioritizing meaningful use cases—like AI automatically transforming meeting notes into tracked action items—drives adoption because it solves real pain points. To that point, employees
estimate using AI helps reduce their workloads by nearly four hours in a typical workweek.
The security conversation is evolving too. An NPU doesn’t just accelerate productivity workloads, it also provides investment protection. Industry analysts have identified edge-based AI security as one of the top use cases for NPU-equipped PCs, with vendors already planning to update endpoint security agents to take advantage of local AI acceleration. That means today’s device refresh lays the foundation for tomorrow’s security innovations.
That said, embedding
security at the firmware and silicon level still matters. A “built-in, not bolted-on” approach safeguards devices from silicon to cloud, blocking threats before they surface in the hardware, firmware, or software and giving IT control without disrupting employees. And by processing AI workloads directly on the device:
- Sensitive data stays local, rather than flowing through unsanctioned public tools.
- Risky tasks, such as opening email attachments, are contained in hardware-enforced micro-virtual machines (micro-VMs), isolating attacks before they spread.
- IT retains visibility and control, with the ability to remotely find, lock, or erase devices even if the PC is turned off when misplaced, lost, or stolen.