AI Workflow Startup Serval Raises $150 Million to Challenge ServiceNow

Serval, a software startup that uses artificial intelligence to automate employee requests and business processes, said Thursday it raised $150 million at a $3 billion valuation. The Sequoia-led financing gives the company more resources to compete with ServiceNow in the systems large employers use to organize internal work.

Its plan contains a revealing tension for the AI industry: selling less manual work still requires substantial human deployment effort. Serval intends to hire more than 500 people over the next year, including over 100 automation engineers in a new professional-services operation called ProServal, according to its announcement.

The financing is Serval’s third round in 12 months. Existing backers including First Round, General Catalyst and Redpoint participated, alongside new investors Conviction and Headline. The valuation is a negotiated price for a private financing, rather than a public-market price or a measure of annual sales.

From writing a workflow to running a business process

Serval’s central automation agent, Catalyst, takes business requirements and turns them into code-based workflows. The company’s product materials describe a system that can work from support-ticket histories, written operating procedures or spreadsheets of tasks, rather than requiring every process to begin with a developer writing an integration.

Consider onboarding an employee. A useful workflow may need to collect information, create access requests, obtain approval and update records in several applications. Generating the code is only one part of that job. The organization must also decide which person can approve access, which records are authoritative and what should happen when an exception occurs.

Those implementation choices explain the proposed services expansion. Serval says roughly half its new automation engineers will be recent college graduates who complete a four-week training program. Their work is meant to help customers identify repetitive processes and redesign them, rather than simply install another chatbot over an existing service desk.

As TTN’s analysis of enterprise AI returns explains, savings depend on the work left for people after the software is deployed. A workflow that requires frequent correction or duplicate record-keeping can move labor around without eliminating much of it. The relevant comparison is the completed process, including supervision and maintenance.

Permissions and testing become part of the product

Catalyst became generally available in August. Serval says it inherits the operator’s permissions, is limited to a team workspace and stages proposed automations as drafts. Authorized users must publish changes before those changes affect employees or production systems. That design separates generating a possible action from granting permission to execute it.

On Wednesday, the company introduced Simulations for testing its Help Desk Agent. Teams can turn historical requests or test data into scenarios, have another model act as the requester and use a judging model to assess the result. Workflows are simulated by default, and access requests do not grant actual access during those tests.

One example in Serval’s explanation concerns a contractor requesting a laptop: the test checks that the request reaches the contractor’s sponsor rather than an ordinary employee’s manager. This illustrates a practical reliability problem. A fluent answer is insufficient when a small difference in employment status changes the correct approval route.

Simulated success still cannot establish that every real-world exception will be handled correctly. For buyers, the value lies in making testing and review repeatable before a change is released. It also connects the product’s commercial case to enterprise data and access controls, where mistakes can carry costs beyond a delayed support ticket.

The incumbent is expanding too

ServiceNow’s results, released Wednesday, show the scale of the competition. It reported $3.988 billion in quarterly subscription revenue, up 21% from a year earlier, and raised its 2026 AI annual contract value target to $1.7 billion. That contract measure differs from revenue already recognized during the quarter.

Serval therefore faces an incumbent selling AI as well as established workflow software. Its funding supplies resources for product development and customer implementation; it does not establish that customers will abandon existing systems at scale. The commercial test is whether a switch produces durable operating savings after migration, permissions, testing and ongoing support are included.