Replacing a mature enterprise platform with LLM-generated code is roughly 10x more expensive once GPU and token costs, engineering rebuild time, and the underlying business-model economics of the model provider are added up.
McDermott walks through the full cost stack: the human capital spent rebuilding what the platform already does, the GPU factory cost behind the model, and the token economics of the model vendor. He argues ServiceNow has 'done the math' and the multiple lands around 10x for a simple application.
saas-vs-ai-economics
Businesses forgive human error but not software error, which preserves a trust advantage for accountable, deterministic platforms over probabilistic language-model output.
McDermott frames this as the practical reason enterprises still pay for a platform rather than self-building: when a language model makes a mistake, there is no one to call, whereas an enterprise platform vendor owns the fix and the relationship.
saas-vs-ai-economics
McDermott's core distinction is 'AI thinks, workflow acts': a language model can instantly recommend the right steps for a business problem, but it cannot execute across the departments, data, and systems needed to actually close the case.
His worked example is a compensation issue that touches sales, HR, finance, legal, and compliance. The LLM answers in milliseconds but doesn't close the case; a workflow platform like ServiceNow is what completes the cross-departmental process.
agentic-enterprise-platforms
ServiceNow's strategy is to be the 'AI control tower' that integrates hyperscalers, language models, and systems of record rather than compete with any of them, mirroring how it responded to earlier fears that hyperscalers would eat software.
McDermott says the same 'why wouldn't the hyperscalers just eat software' fear existed years ago and didn't play out; ServiceNow instead became a major workload driver into those clouds. He expects the same pattern with language models and wants to be the connective fabric across all of them.
agentic-enterprise-platforms
ServiceNow integrated three acquisitions (Moveworks for the agentic front door, Veza for identity, and Armis for OT/IT security) in 20 days, and frames the security move as necessitated by cybercrime being roughly a $1 trillion-a-month problem behind only the US and China as an economic force.
The Armis deal extends ServiceNow from IT into operational technology (networks, manufacturing tools, medical devices, critical infrastructure), aiming to give corporations one automated view across both domains.
agentic-enterprise-platforms
McDermott expects net-new headcount growth to slow sharply at ServiceNow and elsewhere as agents absorb the tactical workload that used to require hiring ahead of growth.
He cites a projection of 2.2 billion AI agents entering the workforce over the next few years, more than the number of new human hires expected, and argues companies will keep humans mainly for relationship-building, judgment calls, and engineering innovation.
ai-workforce-transformation
90% of ServiceNow's own customer service cases are already resolved by agents, with the remaining 10% reserved for human judgment on harder escalations.
McDermott presents this as evidence the lift-and-shift from tactical human work to agent-handled work is already underway inside ServiceNow itself, changing what customer-facing roles actually do day to day.
ai-workforce-transformation
A wave of pandemic-era hiring (2020-2024) is being quietly unwound because much of it happened over Zoom without the scrutiny of in-person hiring, and companies are now using AI to get leaner and lift revenue per employee.
McDermott says this dynamic isn't widely discussed but is a real driver of current headcount discipline: companies over-hired during a crisis period and are now correcting as AI raises the bar for what a given headcount needs to justify.
ai-workforce-transformation
AI adoption is highly uneven by geography and industry: only about 11% of companies in a recent Brazil business audience had moved past the experimentation phase, while US adoption leads globally and industries range from public-sector consolidation to healthcare modernization to financial services already rethinking headcount around AI.
McDermott ties this to a broader claim that most companies know they must 'do something' with AI but are still early; the leaders that have moved to mainstream deployment are treating it as standard modernization rather than an experiment.
ai-workforce-transformation
Enterprise customer conversations have shifted from consultative discovery to demanding fast, prescriptive execution, because customers already assume AI must be embedded and want speed rather than persuasion.
McDermott describes the change as customers no longer wanting a vendor to 'discover their problems' or pitch a solution; they want someone who already understands their business and can deliver a highly predictable, AI-driven result quickly.
agentic-enterprise-platforms
McDermott's founding belief, formed running a deli he bought at 16 with no money (funded by supplier consignment credit), is that a business wins or loses solely on the customer relationship, and he says that idea still drives ServiceNow's culture.
He recounts learning to serve three distinct customer segments (blue-collar workers, senior citizens who wanted delivery, and local kids) and ties the deli's high-touch customer service directly to how he thinks about ServiceNow's platform strategy decades later.
leadership-and-ambition
McDermott argues leadership matters most during turbulent, fast-changing periods, and that AI's purpose is to amplify human ambition and connection rather than replace it.
He uses the framing 'when the tide goes out, you want to be fully dressed' to describe how volatility reveals real leaders, and repeatedly returns to the idea that technology should serve people and relationships, not substitute for them.
leadership-and-ambition
Books referenced
Winners Dream: A Journey from Corner Store to Corner Office - Bill McDermott - McDermott's memoir; he and Guo return to it repeatedly to frame the deli story and his views on leadership, resilience, and ambition.
Media referenced
Blueprint for Agentic Business (ServiceNow white paper) - paper - Internal document McDermott says he uses to teach customer CEOs the difference between a language model and an enterprise platform.
Companies
ServiceNow - McDermott is Chairman and CEO; the platform is positioned as the 'AI control tower' connecting hyperscalers, language models, and systems of record.
SAP - McDermott's prior company, referenced as his other platform-business experience.
Xerox - McDermott's first employer out of the deli; site of the origin story about getting hired by Emerson Fullwood.
Armis - OT/IT security company ServiceNow acquired and integrated in 20 days, extending it into operational technology security.
Moveworks - Acquired by ServiceNow to build the 'agentic front door' to the enterprise.
Veza - Acquired by ServiceNow for human and non-human identity management.
Techniques and frameworks
AI control tower for business reinvention - McDermott's framing for ServiceNow's strategy: sit above and connect hyperscalers, language models, and systems of record rather than compete with them.
AI thinks, workflow acts - McDermott's mental model for why a language model's recommendation is not the same as a workflow platform closing a cross-departmental case.
Health check heuristic - ServiceNow's internal customer-satisfaction-to-system-performance scoring method, mentioned via an employee anecdote.
Summary
Sarah Guo interviews Bill McDermott, Chairman and CEO of ServiceNow and former CEO of SAP, in a wide-ranging conversation that moves from his personal origin story to the economics of enterprise AI to his predictions for corporate headcount. McDermott opens with the story behind his memoir, Winners Dream: buying a deli lease at 16 with no money, funded entirely through supplier consignment credit, and learning that a business survives or dies purely on how well it serves its customers. He carries that lesson into a story about getting hired at Xerox by insisting to a hiring manager that he was walking out with an employee badge that day, and uses both stories to argue that leadership is most valuable during volatile, fast-moving periods like the current AI transition.
The core of the episode is McDermott's rebuttal to the "SaaS apocalypse" thesis, the idea that language models will let enterprises simply generate their own software and stop paying for platforms. His argument rests on cost and accountability: rebuilding what a mature platform already does costs roughly 10x more once GPU and token economics are added to the human capital of rebuilding it, and businesses forgive people for mistakes but not software, so a vendor who owns the fix retains a durable trust advantage. He distills this into a compact mental model, "AI thinks, workflow acts": a model can instantly recommend the right steps to resolve a cross-departmental issue, but only a workflow platform can actually traverse the data and systems to close the case.
McDermott positions ServiceNow's strategy as becoming the "AI control tower" that connects hyperscalers, language models, and systems of record rather than competing with any of them, drawing a direct parallel to years-old fears that hyperscalers would simply absorb software (they didn't, and ServiceNow became a major workload driver into those clouds instead). He describes integrating three recent acquisitions, Moveworks for an agentic front door, Veza for identity, and Armis for operational-technology security, in 20 days, framing the security push around cybercrime being roughly a trillion-dollar-a-month problem trailing only the US and China as an economic force.
On the workforce, McDermott is direct that net-new headcount growth will slow as agents absorb tactical work: he cites a projection of 2.2 billion AI agents entering the global workforce over the next few years and notes that 90% of ServiceNow's own customer service cases are already handled by agents. He also surfaces a less-discussed dynamic, that a wave of 2020-2024 pandemic-era hiring happened over Zoom without normal hiring scrutiny and is now being quietly corrected as AI raises the bar for what headcount has to justify in revenue per employee.
The conversation closes on adoption maturity and customer psychology. McDermott notes that in a recent 800-person Brazil business audience, only about 11% of companies had moved past the AI experimentation phase, with adoption uneven by industry and geography and the US still leading. He says customer conversations have shifted from consultative discovery to demands for fast, prescriptive execution, since customers already assume AI belongs in the solution and want speed rather than persuasion. He ends on a personal note about spending his days across time zones talking directly to quota-carrying reps and customers, and reiterates that AI's role is to serve people and amplify human ambition, not replace it.
Notable Quotes
"People that run businesses understand that people make mistakes. They never will forgive software for making a mistake." - Bill McDermott
"AI thinks. Workflow acts." - Bill McDermott
"When the tide goes out, you want to be fully dressed." - Bill McDermott
"If two people are in the same room at the same time with the same opinion, one of them is redundant." - Bill McDermott