"Doing the same thing with LLMs costs 10 times more"—ServiceNow CEO Bill McDermott on corporate management and competitive advantage in the AI era
ServiceNow CEO Bill McDermott appeared on the No Priors podcast. Known as a "rock star of enterprise technology," he spoke about his consistent philosophy, from his formative experience starting a deli business at age 16 to his career leading Xerox, SAP, and ServiceNow. Of particular note is his clear rebuttal to the "SaaS Apocalypse" theory and his practical perspective on the division of roles between AI and existing platforms.
1. "Customer-centricity" that began with running a deli at age 16
1-1. $5,500 in capital, 500 customers, and 3 segments
"I don't think I did anything special. It was like consolidating several part-time jobs into one," McDermott said. In reality, he integrated three part-time jobs—stocking shelves, working at a gas station, and bussing tables—into running a single deli. The capital was a $5,500 promissory note, which came to $7,000 including interest. It was a simple contract: "If you keep paying, you can own the store. If you are late even once, you lose everything."
The first thing he learned was segment understanding. "Customers like my blue-collar father were rich on Friday and broke by Sunday. Seniors didn't go out and wanted delivery. For the kids, the challenge was getting them to walk past the 7-Eleven a block and a half away." The solution was installing video games, and the kids started saying, "If you want good food, to be treated with respect, and to play games, you come to your store. If you want to steal, you go to 7-Eleven."
"Know your customers and provide what they want in the way they want it. It's a simple equation that applies to any business." This formative experience has become the foundation of his management philosophy at SAP and ServiceNow.
1-2. The anecdote of joining Xerox—"This is not an interview, it's a fight for survival"
During his final interview at Xerox, when told he would be contacted by the hiring manager, McDermott replied, "My father has never broken a promise in 21 years. I promised to bring home an employee badge tonight." The manager at the time broke policy and decided to hire him on the spot, revealing the fact in a speech 25 years later. The weight of the words, "I wouldn't be where I am today without that person's decision," illustrates the relationship between leadership and opportunity.
2. Rebuttal to the SaaS Apocalypse theory—"Trying to replicate it with LLMs costs 10 times more"
2-1. The division of labor theory: "LLMs think, workflows move"
"The cost of trying to replace enterprise platforms in the context of a SaaS apocalypse is enormous," McDermott said. As a concrete estimate, he cited the figure that "trying to replicate a simple application with an LLM costs 10 times more."
He cited three factors for this: implementation costs for platform replacement, human capital costs (the work humans have to redo that the platform used to handle), and GPU factory and token costs.
"Language models provide great advice on handling compensation issues—consider steps 1, 2, and 3. They are excellent. However, they don't close the case," McDermott said. To actually resolve a case, data and workflows spanning multiple departments such as HR, finance, legal, and compliance are required, and that is what a workflow platform like ServiceNow handles.
2-2. The issue of trust: "Software mistakes are not tolerated"
The observation that "people who run businesses understand that humans make mistakes, but they never forgive software for making mistakes" sharply highlights the gap between the probabilistic nature of AI and the certainty and auditability required by enterprises.
The logic is that "LLMs are probably correct, but they are not deterministic. They don't have the company-specific data and context accumulated over decades," which serves as a moat for existing platforms.
3. ServiceNow's "AI Control Tower" strategy
3-1. A "fabric" that bundles LLMs, hyperscalers, and systems of record
"We are trying to become the AI control tower for business transformation. We integrate with every hyperscaler, every language model, and every system of record," McDermott said. Currently, ServiceNow has integrated approximately 800 systems of record, with 8.5 billion workflows running and 7 trillion transactions processed on the platform.
This positioning overlaps with history, where there was once a concern that "hyperscalers would erode software," but it never materialized. "The same thing was said back then. That hyperscalers would absorb all workloads. They are great companies, but each is good at different things. ServiceNow is flowing billions of workloads to the cloud behind them as well." This empirical rule is the basis for his confidence that the same logic will hold true in the LLM era.
3-2. Expansion into security—"Cybercrime is the world's third-largest economy"
Another strategic move ServiceNow has recently made is entering the security sector. With the acquisition of Armis, they are advancing the integration of security management, including OT (Operational Technology). Adopting a scale where "the U.S. is the world's largest economy, China is second, and cybercrime is third—a $1 trillion-a-month problem," they are taking an approach that provides management and visibility while integrating with existing security platforms.
The vision of managing the "physical world and digital world," including manufacturing equipment, medical devices, and critical infrastructure, on a single AI platform leverages ServiceNow's cross-functional position. Even more noteworthy is their execution capability, having "completed these integrations in 20 days," which demonstrates the company's high level of engineering prowess.
4. AI agents are changing corporate organizations—"Growth without increasing headcount is now possible"
4-1. 90% of customer service to be handled by AI agents
"Now, 90% of our customer service cases are handled by agents. Humans are involved in only 10%," said Mr. McDermott. This is a "shift and transformation of work," a change where humans can focus on high-level judgment and building human relationships.
4-2. The outlook of "growing without increasing headcount in five years"
He presented the outlook that "in a growth company like ServiceNow, we previously had to hire thousands of people as our finance, HR, and support functions grew. Now, agents can handle much of that." His forecast is that "future net headcount growth will decrease dramatically. However, the company's productivity will improve significantly."
At the same time, he provided a framework: "It is important to distinguish between what AI can and should do, and what humans should do. Building loyalty, trust, promises and their fulfillment, and the long-term value of human relationships—these will remain in the human domain."
5. "Which SaaS is dangerous and which is safe?"—Conditions for vulnerability
5-1. The difference between platforms and point solutions
"Theoretically, SaaS companies with data and context should be more valuable. However, if there is a vulnerability, it lies in single-function companies that are not cross-departmental," said Mr. McDermott.
The judgment criteria that "critical systems of record that span multiple departments and are difficult to replicate are safe. But if it provides only one function, has low added value, and is not at the top of the CEO's priority list, the vulnerability equation rises rapidly" is a perspective that can be directly applied to investment evaluation of SaaS companies.
Specifically, this analysis aligns with the assessment that "platform types like CrowdStrike are more likely to have immunity. On the other hand, single-function point solutions are high risk."
5-3. The current state of AI adoption—"More companies are moving past the experimental phase"
The figure that "even in a meeting last night in Brazil, only 11% of companies have moved beyond the experimental phase" shows that AI adoption varies significantly by geography and industry. Financial services are moving the fastest, the public sector is in the integration phase of complexity, and healthcare is in the modernization phase. He noted that the U.S. leads the world in AI consumption and adoption.
Conclusion
The logic presented by Bill McDermott is simple for investors. SaaS companies that are "platform-based, cross-functional, and have high switching costs," including ServiceNow, will be strengthened by the rise of AI. Conversely, point solutions that are "single-function, department-limited, and have low replacement costs" face higher risks. The division of labor theory that "LLMs think, workflows move" provides a practical framework for the coexistence of AI and SaaS. The figures of $13 billion in annual revenue and a 20% growth rate are also a realistic validation of this logic.