Enterprise software has long operated on a relatively stable hierarchy of power: The companies that owned the interface largely owned the customer relationship. Employees moved through dashboards, tabs, forms, and menus; software vendors sold more seats, expanded across departments, and steadily compounded recurring revenue.
Agentic AI is beginning to destabilize that model. Increasingly, enterprise users no longer need to navigate software directly to complete routine work. AI agents can coordinate actions across multiple systems through natural-language commands alone.
That possibility has rattled the software industry. Earlier this year, SaaS stocks sold off sharply as investors questioned whether AI agents could weaken sticky interfaces, compress seat growth, and erode the economics that powered enterprise software for decades.
The question now hanging over the industry is whether AI agents will hollow out enterprise software altogether, or if they’ll reorder where value accrues within it. Few executives have pushed back against that first narrative more aggressively than Bill McDermott, the longtime ServiceNow CEO who argues investors fundamentally misunderstand how enterprise AI will actually get deployed inside large organizations.
“AI thinks,” McDermott tells Fast Company. “It’s got tremendous compute power. But it doesn’t act.”
That distinction sits at the center of ServiceNow’s broader AI strategy. While many investors worry that hyperscalers and foundation model companies will eventually absorb large portions of enterprise software, McDermott argues the rise of AI is actually increasing the importance of orchestration, workflow governance, and operational execution systems.
“When you’re running a company, and you want the digital agents to work with the humans, or even in a lot of cases do the work that the humans are doing, they just have to execute along the lines of the business process so things actually get done,” he says.
So far, investors’ fears around AI disruption have not materially slowed ServiceNow’s growth. The company still expects more than $15.7 billion in subscription revenue in 2026, while its Now Assist AI business reached $750 million in annual contract value in Q1 and is targeting $1.5 billion by year’s end.
The company argues AI adoption is deepening customer reliance on the platform rather than weakening it. According to ServiceNow, 91% of net-new annual contract value in 2025 came from deals involving five or more products, while Now Assist deals that included three or more products grew nearly 70% year over year.
Why operational reality still slows AI disruption
Daniel Newman, CEO and chief analyst at Futurum Research Group, says the current AI cycle is moving faster than any prior enterprise technology transition, but many investors initially underestimated the operational inertia built into large organizations.
“The deepest moat that’s making transformation and change to new technologies much harder is merely that humans change much slower than technology,” Newman says.
That operational reality has become central to how many incumbent software companies now defend themselves against AI disruption. While Silicon Valley increasingly imagines autonomous AI systems replacing large swaths of enterprise software, companies still face governance, compliance, security, transaction, and data privacy constraints that remain difficult to solve at enterprise scale.
McDermott argues much of the market still misunderstands the complexity involved once companies move beyond chatbot experimentation and into real operational systems.
“You can’t just say, ‘Here’s the model. Good luck running your company with it,’” he says. McDermott also rejects the assumption that hyperscalers will inevitably absorb the enterprise execution layer entirely.
“These are significant companies with billions invested in infrastructure,” McDermott says of the major cloud providers. “By spreading our wings and using those clouds, it actually alleviates concerns because this is about expansion in the global economy.”
Still, parts of the original SaaSpocalypse thesis are already beginning to materialize. “I do think there’s going to be seat compression,” Newman says. “I think if you had 100 seats before AI, maybe you have 80 now.”
The larger shift, he argues, is economic. “Seat-based pricing is dead,” Newman says. “It’s not going to be about the number of users anymore. It’s going to be about agents deployed and tokens consumed.”
The AI stack is reordering in real time
That transition is already forcing many enterprise software companies to rethink how they package and monetize AI products. Workday cut 8.5% of its workforce earlier this year amid broader AI restructuring pressures. Atlassian recently reported its first-ever enterprise seat-count decline, intensifying concerns that AI-native coordination systems could weaken traditional collaboration platforms.
McDermott says ServiceNow is already navigating that transition internally. Roughly half the company’s revenue now comes from consumption-based pricing, though he argues customers still prefer hybrid models blending seats with usage-based billing.
“That’s the Goldilocks formula,” he says.
The new enterprise moat may be workflow depth
Pat Casey, ServiceNow’s CTO and one of the company’s earliest cofounders, argues ServiceNow’s AI positioning stems from architectural decisions made long before generative AI emerged.
“I think if you look at the way most of the industry is applying AI, it’s what I’d call a horizontal assistant layer,” Casey says.
ServiceNow, by contrast, increasingly positions itself as the operational layer sitting beneath enterprise AI systems. Casey argues many systems of record were never designed to be orchestrated externally once real-world business processes begin cascading across approvals, sourcing, compliance, and downstream operational dependencies.
A FedEx demonstration shown during ServiceNow’s Knowledge 2026 conference became one of the clearest illustrations of that thesis.
In the demo, ServiceNow Otto, the company’s unified AI interface combining Moveworks and Now Assist, helped a FedEx distribution manager assess whether the company was prepared for a Mother’s Day shipping surge. Otto reviewed staffing coverage, identified a 37-person staffing gap, generated a compliant hiring requisition, scheduled interviews, and deployed an AI agent to manage follow-ups.
“Putting the request in is maybe one percent of the work,” Casey says. “The actual delivery process—that’s where the complexity and the bodies are buried.”
The hyperscaler threat may be bigger than SaaS companies admit
That operational complexity increasingly sits at the center of the enterprise AI battle. Historically, infrastructure companies rarely stop expanding upward once they control distribution. Nearly every layer of the enterprise stack now risks abstraction from above. Data platforms remain central to enterprise AI development. Workflow platforms increasingly control operational execution. Governance layers may emerge as critical trust infrastructure.
At the same time, open interoperability standards and emerging multi-agent protocols could eventually commoditize portions of the orchestration layer itself.
As enterprises deploy larger fleets of AI agents, governance has emerged as one of ServiceNow’s biggest strategic priorities. The company’s expanded AI Control Tower strategy aims to govern AI models, agents, workflows, and datasets across environments spanning Microsoft, Google, Amazon Web Services, OpenAI, Anthropic, and third-party systems.
“There’s always the risk somebody put a Mac Mini in a closet, installed an open-source model on it, and now it’s quietly doing HR tasks nobody authorized,” Casey says.
For now, the enterprise software market appears caught between competing possibilities: foundation model companies expanding upward into enterprise execution, hyperscalers consolidating infrastructure power, or incumbent workflow platforms using operational complexity itself as the next moat.
Newman believes many foundation model companies are already discovering how difficult enterprise deployment becomes in practice. “That’s why OpenAI is partnering with consulting companies and Anthropic has partnered with every one of these enterprise software companies,” he says.
Still, he believes not every SaaS company survives the transition equally well.
“The companies I worry about most are what I call features masquerading as companies,” Newman says, pointing toward lightweight productivity and collaboration platforms. “The Notions, the Mondays, the ClickUps. Those are the companies I think face the most pressure because they’re really productivity wrappers around human coordination.”
