Key Takeaways

Raw Models are the New Utility. Context is Your Moat.

Harry Stebbings’ conversation with Nikash Arora, CEO of Palo Alto Networks, cuts through the AI hype. Forget chasing the latest, flashiest model. Arora’s core insight: the raw intelligence of AI models is rapidly becoming a commodity. As models get better and better, their distinctions will matter less and less. You won't rip out one for another based on a marginal improvement in raw output. The real game-changer for enterprise AI isn't the model itself, but the specific, proprietary 'context' an organization builds around it. This means your historical data, your unique customer interactions, and your codified domain knowledge. It's the secret sauce that takes a generalist AI and makes it a specialist that truly understands your business problems.

Domain Knowledge is Now Code

Arora sees a near future, within the next five years, where domain expertise holds as much weight as raw model intelligence. Imagine your top performers – the ones who consistently solve complex problems. Their hidden knowledge, their "playbooks" for tricky situations, that's what needs to be extracted and codified. This isn’t just about feeding data to an AI; it’s about creating a living, breathing 'learning system' within your organization. "As I build that organizational instead of context," Arora explains, “then I can stick any model I want on it. And the model distinction will not matter because the context will become as important or perhaps more important.” This 'context' transforms generic AI into an invaluable asset, ensuring it delivers reliable, actionable results specific to your operations.

The Overlooked Engine: Training Data

For all the talk about neural networks and foundation models, Arora points to a more mundane, yet critical, piece of the puzzle: training data. “I think the enterprises ability to absorb this or digest this or perhaps use this to their advantage depends on their ability to create training data as fast as they can and I think not enough people are focused on training data,” he warns. This isn't just about having a lot of data; it's about actively cultivating and transforming every customer interaction, every support call, every problem solved into a learning opportunity. This process, often overlooked, is what allows an AI to move from a respectable 70% accuracy to a rock-solid 99%, making it truly dependable for critical business functions. It means abstracting and codifying the solutions from human experts into rules and playbooks your AI can learn from.

What to Do With This

Don't wait for a better foundation model. Instead, identify your top 3 most complex, recurring business problems and begin extracting the "human context" around them. Interview your best problem-solvers to codify their decision trees and unique knowledge into lightweight internal playbooks or knowledge bases, tagging them with specific domain terms. Then, direct every customer support interaction, every sales call, or every engineering fix to generate structured training data, explicitly linking the problem to your newly codified solutions.