ServiceNow Builds Startup Inside Itself to Fight AI-Native Rivals

ServiceNow says its new conversational AI service desk, Flow, can be set up within a day with "no implementation project, CMDB migration or infrastructure required" — a direct answer to the two complaints rivals have spent two years building businesses on: implementations take too long, and employees want help inside tools they already use.

Flow runs inside Slack and Microsoft Teams. Employees ask for a password reset or application access in plain language, and ServiceNow says agentic AI can handle the request or escalate it when needed. The product remains in controlled availability.

The competitive pressure is concrete. Serval, an IT service management startup that says it intends to replace ServiceNow, reached a $1 billion valuation last December after raising $75 million in a round led by Sequoia Capital. Sequoia partner Anas Biad said the last time the firm saw customer feedback that strong was when it backed ServiceNow itself 16 years earlier.

Salesforce CEO Marc Benioff has also questioned ServiceNow's reach. He has pointed out that ServiceNow automates work for roughly 9,000 companies, while Slack — owned by Salesforce — already sits inside a million.

"For too long, small and medium businesses had to choose between simplicity and scale. Flow gives them both," Bill McDermott, ServiceNow's chairman and CEO, told Fast Company in an email exchange. "We created Flow to give the Fortune 500,000 the power of enterprise AI without the enterprise complexity — a conversational service desk delivering value from day one. Zero upfront expense. Pay only for what you consume."

McDermott argues ServiceNow's enterprise workflow experience separates Flow from easier-to-build competitors. "Anyone can vibe code a conversation. Very few can engineer an outcome," he said. "Without orchestration, context, security and trust, you're creating vibe slop that will flop."

How was Flow built?

Amit Zavery, ServiceNow's president, CPO and COO, says Flow grew out of requests from leaner IT teams that wanted its automation without a full-scale implementation. "They want to get started immediately, without going through any kind of configuration process. They want to start using a product, see where it goes, and pay as they go. That's a consumption-oriented mindset," Zavery told Fast Company.

Zavery assembled a small team with backgrounds in IT service management and AI development tools, then gave its members a founder-like mandate. "I said, you have no restrictions. It was like a well-funded startup, in a way," he said.

ServiceNow says the team built Flow in three months. The product also draws on years of knowledge about how IT departments handle employee problems, which requests can be automated, and when a workflow needs to reach another system or person.

Keith Kirkpatrick, vice president of research at Futurum, sees Flow as part of a broader shift forcing established vendors to package expertise in products that are faster to adopt and simpler to buy. "Incumbent software vendors will need to pivot to these types of more nimble offerings in order to fend off challenges from other vendors, as well as in-house development using AI tools," he said.

Can Flow outrun AI-native rivals?

Serval says it automates more than half its customers' IT tickets, and some customers have replaced incumbent systems entirely. Serval cofounder and CEO Jake Stauch previously claimed some ServiceNow customers deployed less than 10% of the ServiceNow AI products they purchased.

Salesforce said more than 180 organizations had chosen Agentforce IT Service four months after general availability. Flow can meet employees in the same channels, but ServiceNow does not control those channels or their commercial terms.

Zavery argues competitors cannot quickly reproduce ServiceNow's operating knowledge. "AI has now made it much easier to write code, but simply being able to write the code is not what matters," he said. "Knowing what to ask the system to write, and knowing whether what you are building actually makes sense, still requires you to understand what you are doing."

He points to scale as evidence: "I don't think there is another company today that can do what we do, at the level we can do it. We run 8 trillion transactions on ServiceNow on a yearly basis."

What does Flow change for buyers?

Flow can turn recurring requests into automations, Zavery says. "It gives you the ability to simply say, 'slash automate,' and the system creates the entire automation for you. Essentially, the next time that same request comes through, no human needs to interact with it on the back end."

ServiceNow says Flow includes guardrails to prevent AI agents from accessing systems or taking actions without authorization, while its separate AI Control Tower provides broader oversight. "Flow is not designed to manage your entire AI estate. That is where AI Control Tower comes in," Zavery said. He did not specify mechanisms to reverse or contain a harmful agentic action after execution.

Flow sits alongside overlapping ServiceNow products. The company spent $2.85 billion on Moveworks, another conversational entry point for employee requests, and already offers EmployeeWorks and Otto. Zavery says EmployeeWorks and Otto will continue serving larger enterprises.

Early customers include Serenity EHS, which already builds on ServiceNow's platform, and the U.S. Navy's Fleet Numerical Meteorology and Oceanography Center, which believes Flow could let employees build repeatable workflows inside Microsoft Teams.

What are the economics?

ServiceNow expects Flow to cut ticket volume by 40% and let teams build automations in five minutes. The company has clarified those figures are "expectations" while the product remains in controlled availability — they have yet to be demonstrated at scale.

ServiceNow's AI annual contract value exceeded $1 billion in the second quarter, and 658 customers each generated more than $5 million in annual contract value. Flow adds a different buying path: credit card sign-up for new customers and consumption-based use for existing customers whose plans include AI.

Futurum's Kirkpatrick argues ServiceNow will need that flexibility. "A smaller company with basic support needs may be happy with an AI vendor, particularly if all they are trying to do is set up a basic support system," he said.

Winning smaller companies with lighter deployments could mean smaller contracts, even if some expand over time. The open question is whether ServiceNow can build that lower-commitment business without weakening the high-value enterprise model it already depends on — or whether the simpler product proves sufficient even as customers grow.