Research · November 7, 2026

Serval's $150M Series C — Replacing the System of Record, Not Adding Another AI Layer

Serval raised $150 million in a Series C round led by Sequoia at a $3 billion valuation to expand Catalyst, its automation agent, and build a modern system of record — replacing platforms like ServiceNow rather than adding another AI layer on top of one.

What happened

On October 29, 2026, Serval, a San Francisco-based enterprise automation company, announced a $150 million Series C at a $3 billion valuation, led by Sequoia. Elad Gil, Evantic, First Round, General Catalyst, and Redpoint participated, with Conviction and Headline joining as new investors. It is the company's third funding round in 12 months, and Reuters reports total funding raised to date of $277 million. The prior round, a $75 million Series B announced December 11, 2025, was also led by Sequoia at a $1 billion valuation — so the new valuation is roughly triple the December mark in under a year.

Co-founders Jake Stauch (CEO) and Alex McLeod (CTO) frame the mission as extending engineers' building power to everyone inside large enterprises. Their pitch starts from an observation that AI coding tools transformed engineering work first, while the rest of the enterprise still waits for the same transformation — and that the bottleneck is not a lack of AI models, but who is allowed to build with them.

What Catalyst actually does

Serval's automation agent is called Catalyst. With it, employees across IT, security, HR, finance, legal, and other departments build and run code-based workflows without engineering experience: they bring the domain expertise, and Catalyst translates business requirements into deterministic tools that run under strict enterprise controls, with guardrails, permission checks, and approvals managed by administrators.

The important design choice is where those automations live. Catalyst is embedded in Serval's full enterprise service management platform, so teams build and run automations in the same place they track requests, manage assets, and orchestrate changes and approvals. In the earlier Series B announcement, the company described the underlying model in the same terms: a team outlines a process in natural language, the agent produces, tests, and publishes the full automation, and the team can then refine it with further natural-language prompts or directly at the code level. The company says the resulting workflows are explainable, auditable, permissioned, and fully traceable.

That "same place" claim is what separates Serval from the point-solution agent pattern. A dispute-resolution agent or a procurement-exception agent automates one workflow inside someone else's system. Serval wants to be the system itself — the record where requests, assets, and approvals already live — so the automation does not need to be bolted onto an incumbent's database.

The replacement claim, and its one named proof point

Serval's sharpest claim is that it is not an AI layer on top of a system of record but a replacement for one. "It's not an additional layer of AI spend on top of your existing system of record, it is a true alternative replacement for platforms like ServiceNow," Stauch told Reuters. Because it fully replaces the system of record, the company argues, customers consolidate spending rather than adding another budget line — a direct answer to the "SaaSpocalypse" anxiety that AI tools merely stack new subscriptions onto entrenched vendors like Workday, Salesforce, and SAP.

The migration proof point the company offers is a single automotive manufacturer that replaced ServiceNow with Serval in 45 days, company-wide, spanning IT service management, the configuration management database, hardware and software asset management, and a long tail of departmental workflows. The company announcement leaves the manufacturer anonymous; Reuters identifies it as Lucid Motors. One migration, however fast, is not a pattern — but it is the kind of claim incumbents cannot ignore if it repeats, because a 45-day rip-and-replace undercuts the usual argument that switching costs protect the incumbent.

The company also reports traction in its own terms: annual recurring revenue doubling every quarter since the Series B, two Fortune 20 companies onboarded in the last six months, and Fox Corporation, WPP, Live Nation, and Ramp now building with Serval. No absolute revenue figure, customer count, or retention number is disclosed, so those claims can be read only as direction, not scale.

What the money funds: 500 hires and a services arm

Serval says it will hire more than 500 people over the next 12 months. More than 100 of those hires will be Automation Engineers assigned to ProServal, a new professional services arm whose staff work alongside enterprise teams to spot automation opportunities, rework existing processes, and remove repetitive work. Roughly half of those engineers will be recent college graduates completing a four-week Serval training program — a deliberate bet that the generation entering the workforce already fluent in AI can be placed inside complex organizations, shoulder to shoulder with the people who know how those businesses actually run.

That services arm is worth noting because it cuts against the pure-software story. If the product truly let any employee build workflows with no engineering experience, a 100-plus-person deployment team would be less necessary. ProServal suggests the company knows the hard part is not generating a workflow but redesigning the broken process underneath it — work that still requires people inside the customer's building.

The founders bring a relevant pedigree: Stauch and McLeod previously led product and engineering at Verkada, where, according to Sequoia's account of the partnership, IT customers repeatedly asked the pair to fix their helpdesk. Sequoia said the last time it had encountered comparable customer feedback about a system of record enabling horizontal enterprise automation was 16 years earlier, when it partnered with ServiceNow — and summed up its thesis as "Serval is not AI for ITSM — it is ITSM built from AI."

Why this is a different bet from the agent builders

This queue has recently filled with companies selling agents that do a specific job — finance disputes, procurement exceptions, benefits questions, wealth operations, home-services calls. Serval sells something one layer up: not an agent that performs a workflow, but a platform where non-engineers build the workflow themselves, inside a system of record meant to displace the incumbent's. If AI agents keep being limited less by model capability than by who controls the record and the permissions around it, then the scarce asset may be the system of record itself — and the company that owns it gets to decide which automations are allowed to run.

What this does not show

The announcement discloses no absolute revenue, customer count, retention, or pricing figure — ARR "doubling every quarter" is a growth rate without a base, and the two Fortune 20 onboardings are unnamed. Only one migration is described in any detail, and its identification as Lucid comes from Reuters, not from Serval's own announcement. There is no disclosed evidence on how many Serval-built automations survive contact with real enterprise change — reorganizations, policy rewrites, system migrations — or on what share of workflows still requires human approval for sensitive actions, beyond the company's statement that IT and security teams control permissions and that customers generally still want human sign-off for password resets, application access, and changes to sensitive systems. The hiring plan is a plan funded by this round, not a reported result.

What would turn this round from a valuation story into a displacement story is repetition: a second and third 45-day migration, named, and evidence that the automations built by non-engineers keep working as the enterprise around them changes. Until then, Serval's case rests on a plausible inversion — that the system of record, not the agent, is the moat — and an unproven claim that employees who have never written code can build durable enterprise software with an agent's help.

Sources