While venture capital pours billions into replacing human jobs with AI, BlackBerry just explained why their software can't be replaced—and what that tells us about where the automation economy hits a wall.
The Summary
- BlackBerry CEO John Giamatteo says AI can't displace his company's safety-certified software that powers critical systems in vehicles, medical devices, and industrial automation.
- BlackBerry's QNX operating system now runs in over 275 million vehicles globally, with expanding deployments in robotics and medical devices where certification requirements create moats.
- The company has evolved from smartphone maker to infrastructure software provider focused on safety-critical systems where regulatory approval, not code speed, determines market access.
- This represents a rare category of B2B software where AI tools enhance but can't replace the underlying certified platform.
The Signal
BlackBerry's transformation reveals a crucial distinction most software companies are ignoring: not all code is created equal. While AI agents are rapidly automating routine software development and threatening SaaS margins, safety-certified software operates under entirely different rules. You can't ship a medical device OS update because Claude wrote elegant code. You need certification bodies, compliance trails, and liability frameworks that take years to navigate.
QNX's footprint of 275 million vehicles represents embedded infrastructure that compounds in value as autonomy increases. More self-driving features mean more safety-critical software, not less. The same pattern plays out in robotics and industrial automation where BlackBerry is expanding. Every warehouse robot that works alongside humans, every surgical device that touches patients, every factory system that could hurt someone requires certified software foundations.
"AI isn't ready to displace safety-certified products."
This isn't Luddite resistance to automation. It's recognition that certification creates architectural lock-in. When an automaker spends five years getting QNX certified for their vehicle platform, they're not ripping it out because a new AI model writes better C code. The switching costs aren't technical, they're regulatory and temporal. You can't fast-follow your way into FDA or ISO 26262 approval.
What makes this particularly relevant for the agent economy: the infrastructure layer resists automation differently than the application layer. Consider the stack:
- Bottom: Safety-certified OS (QNX) — AI-resistant, human-certified, decade-long replacement cycles
- Middle: Control systems and middleware — Partially automatable, still needs human oversight for critical paths
- Top: User-facing applications and interfaces — Rapidly automatable, where AI agents live
The agent revolution concentrates at the top of the stack. But agents need certified foundations to operate in physical spaces. Your warehouse robot's planning algorithm might be pure AI, but it's running on certified real-time operating systems that took years to approve. The economics matter: BlackBerry doesn't compete on features or fast iteration. They compete on the compound value of installed, certified infrastructure that becomes more valuable as more automation layers stack on top.
The Implication
If you're building AI agents for physical environments, understand where certification requirements create choke points. The companies that control safety-certified infrastructure platforms have pricing power and customer lock-in that pure software players don't. BlackBerry's bet is that the agent economy needs certified foundations more than ever. They're probably right. Watch for similar dynamics in medical AI, industrial automation, and autonomous vehicles. The most valuable infrastructure won't be the most innovative. It will be the most certified. And certification timelines don't compress with better models.