‘Frontier AI companies have no moat’: ServiceNow’s Amit Zavery on why markets are getting software wrong

ServiceNow’s Amit Zavery says investors are underestimating how AI will reshape software, arguing that frontier AI models lack a durable moat while enterprise context, domain expertise and measurable outcomes will remain critical.

By Shereen Bhan

(Photo Credit : AI Generated )

ServiceNow’s President, Chief Product Officer and Chief Operating Officer Amit Zavery believes investors are making a fundamental mistake by putting all software companies in the same basket as artificial intelligence reshapes the industry.

In an interview with CNBC-TV18 at ServiceNow’s World Forum in Mumbai, Zavery argued that frontier AI companies do not currently have a durable moat, while enterprise software platforms with deep domain knowledge and years of operational context could be harder to displace.

“Not in the current model, no,” Zavery said when asked whether frontier AI companies have a moat. He likened large language models to infrastructure such as chips, arguing that ServiceNow can switch between models depending on their cost and performance.

“Customers don’t care what model I use; they care about the outcome,” he said. “I keep changing models based on what works.”

That flexibility is central to Zavery’s argument that the market is failing to distinguish between companies that AI threatens and those that can use it to strengthen their businesses.

ServiceNow has raised its expectations for AI-related annual contract value from $1 billion to $1.5 billion, Zavery said, arguing that AI is accelerating demand rather than eroding the company’s business. He described ServiceNow as an “operating system” connecting different systems inside an enterprise, making it difficult to replace.

The company’s advantage, according to Zavery, lies less in the underlying AI model and more in the enterprise context built around it.

“If you look at the IP inside my product today, less than 10% is the LLM,” he said. The rest, he argued, comes from years of understanding workflows, business requirements, governance and why decisions were made. ServiceNow currently runs 100 billion workflows and 8 trillion transactions a year, according to Zavery.

That context also explains why he does not see partnerships with OpenAI and Anthropic as a threat. Frontier model companies provide the intelligence layer, he said, but they lack the enterprise execution layer and do not have access to ServiceNow’s customer data.

Zavery also rejected the idea that AI makes software development alone the key competitive advantage. Writing software may become cheaper, but maintaining enterprise software remains the harder problem.

“Writing the software costs 20% for an enterprise; maintaining it is the other 80%,” he said, pointing to governance, security, compliance, upgrades and maintenance as areas where domain expertise remains critical.

The disruption, however, will be significant. Zavery expects software development to become faster and pricing models to move beyond traditional seat-based licensing. He said 50% of ServiceNow’s net-new revenue over the past year was non-seat-based, with customers increasingly paying according to outcomes, consumption or service-level agreements.

He also expects AI to reshape jobs rather than simply eliminate them. At ServiceNow, he said, one person can increasingly manage multiple AI agents performing software-related tasks.

“I used to measure people by the number of lines of code. Now I’m measuring how many agents they manage,” Zavery said.

For Zavery, the biggest challenge for ServiceNow is therefore not the existence of AI competitors, but keeping pace with an unprecedented rate of technological change.

“I don’t think I’ve ever seen this speed of change,” he said. “Cloud took 10 years... What we did with cloud in 10 years is happening in one month in this software industry.”

His message to investors is straightforward: the winners in enterprise software will not necessarily be those building the most powerful AI models, but those that can combine AI with deep context, customer relationships and measurable business outcomes.