Palo Alto Networks CEO Nikesh Arora says 90% of enterprise workers lack AI skills

Palo Alto Networks CEO Nikesh Arora thinks companies are walking into the next phase of work unprepared. On the “20VC” podcast, Arora said the “challenge right now” is that “90% of the enterprise employees are not AI savvy.” And crucially, he does not believe there is some magic training program that can fix it fast enough.

Arora framed the moment as something closer to natural selection than upskilling. He said there is “no training course he can send his 21,000 employees” to, so employees have to “learn on their own.” That lack of a ready-made curriculum is why he called it “a Darwinian moment where everybody has to figure out who's really good.” For a company with a total market cap over $235 billion, the implication is clear: if the workforce cannot prove it can use and build with AI, roles will either shrink or disappear.

This is not just a workforce story. It is also an org design story, and it is happening while AI tools keep getting cheaper and more capable, especially for tasks that used to require dedicated headcount. Arora said other companies are facing the same reality and choosing to respond with mass layoffs. He referenced examples from outside cybersecurity, pointing to Coinbase CEO Brian Armstrong and Block CEO Jack Dorsey, both of whom publicly discussed rebuilding with fewer people.

Arora said Armstrong and Dorsey essentially concluded that there was “no redemption,” because they could not train everyone to perform in an AI-shaped environment. His description was that they would “decimate” their organizations and “start building from scratch,” cutting roughly 30% to 40% of headcount to find the people who can do the work now. The source notes Dorsey announced in February that Block was laying off over 4,000 workers, nearly half of the company. It also says that in May, Coinbase announced it was laying off 14% of its workforce, affecting about 700 roles, while Armstrong wrote that the cuts were designed to make the crypto firm “leaner, faster, and more efficient for our next phase of growth.”

Palo Alto Networks, though, is taking a different route. Arora said his company is using natural attrition instead of large-scale layoffs, gradually replacing workers as it grows the technical bench it believes the AI moment requires. He said Palo Alto Networks knows where to find future technical talent, and he pointed to how the company hires: “We’ve been hiring people only through hackathons.” The timeframe he gave is telling. He said, “Give me 12 months, I’ll have sort of transformed 20, 25% of my team,” and “Give me three years, I'll have hopefully enough AI savvy people working at Palo Alto.”

That matters because Palo Alto Networks is not retrenching. The source says the company “has added 5,423 total employees to its headcount from the end of fiscal 2025 to the third quarter of 2026,” according to its most recent 10-Q filing. So the operational bet is: rather than freeze hiring or cut across the board, keep expanding while changing the mix, using attrition and selective pipelines to shift toward AI-capable roles.

Arora also drew lines around the functions he thinks AI will compress first. He questioned why he needs “400, 600 people in marketing” when frontier models can be trained on marketing strategies and a company’s specific voice. His point is not that marketing gets automated into irrelevance overnight. It is that scale labor may be less valuable than consistent execution quality. He said his “biggest problem in marketing” is having “600 people” but not all of them fully understanding how to consistently deliver his tone of voice, value proposition, and “how not to break my brand.”

His “rule of thumb” was that in the next three years companies will “probably have half of the people” in general and administrative roles like marketing, HR, and finance. He linked that to the idea that AI applications will soon do more than draft text. In his view, these tools will provide feedback and effectively act as opinionated intermediaries. He described an “AI assistant” that reviews copy and tells a user that it “sucks,” that it is “not good enough,” and what to recommend to fix it. He said “This has an opinion,” and that would make the average employee “much smarter than they were today.” The second-order effect is brutal: if AI does more of the work, then fewer people may be needed because the tool is already doing part of the job.

But Arora says the endpoint is not a pure headcount collapse. He said he wants more technical and sales resources, and he previously said he wants more cybersecurity engineers and researchers. During the interview, he also said some employees want more AI resources to help implement plans to transform marketing and HR. That is his rebuttal to the common fear that AI will automatically mean fewer people working. He said he believes that “what's going to happen is you can't imagine the number of people on my team who want more technical resources, more AI savvy resources because they want to do exactly these things.”

For executives at other cybersecurity and software companies, the stakes are straightforward. If you are hiring or budgeting for the next three years, you are not just buying headcount. You are buying organizational adaptability. Arora’s message is that the training problem is not theoretical, it is operational. Without a scalable way to make the existing workforce “AI savvy,” companies either reduce general and administrative layers quickly, or they replace them gradually while hiring for technical capability today.