What it is
Serval is an AI-native enterprise service management platform: a ticketing system, a help desk agent that answers employees in Slack, Microsoft Teams, email, phone or a portal, and a builder agent that writes the automations the help desk agent is allowed to run. It started in IT and now sells into HR, security, finance and legal requests: password resets and device issues, but also PTO bookings in the HRIS, NDA and DocuSign requests, purchase-order lookups and revoking unused SaaS seats.
The design choice that matters is the split between two agents. Catalyst, the admin-side agent (generally available since August 2026), turns a plain-language request (“when someone asks for Figma, check their manager’s approval and provision through SCIM”) into a deterministic TypeScript workflow that an admin reviews and publishes. The help desk agent can act only through those published workflows, with a full audit trail; Serval states that permissions are “not subject to LLM judgement.” That is a better answer to “what can the bot actually do to our systems” than most AI service desks give.
Around the two agents sit the ITSM basics: ticketing with SLAs and assignment rules, ITIL request, incident, problem and change management, a knowledge base, a service catalog, an asset database and CMDB, and access management with just-in-time grants, automatic deprovisioning and user access reviews. Roles and groups import from Okta, Microsoft Entra or Google Workspace. Outside systems connect over API, MCP and A2A. It runs as a full replacement or as a layer with two-way sync to ServiceNow or Jira Service Management.
Company status: founded in 2024 by Jake Stauch and Alex McLeod, who met at Verkada. On 2026-10-29 it raised a $150M Series C led by Sequoia at a $3B post-money valuation, its third round in 12 months, for $277M raised in total (Fortune). It says ARR has doubled every quarter since the December 2025 Series B but discloses no dollar figure. Headcount went from about 20 in January 2026 to about 200. It is the acquirer, not a target: it bought the supply-chain security startup Ensignia in October 2026.
Pricing reality
Serval’s pricing page lists no prices. Every deal starts with a paid engagement in four phases (plan, pilot, deploy, go-live) staffed by Serval’s own deployment engineers, then a single platform fee. Third-party write-ups from earlier in 2026 describe a guarantee of 50% of IT tickets automated by the end of the pilot; the pricing page as updated on 2026-10-29 no longer states it, so ask for it in writing rather than assuming it.
The only public number is on AWS Marketplace: one “Serval Platform” dimension at $500,000 per 12 months, marked “Contact for Private Offer,” with usage above the contract billed separately and no refunds. Read it as an enterprise list anchor, not a quote; we found no verified negotiated contract. For scale, ServiceNow’s own entry puts 50 fulfillers at roughly $48,000 to $90,000 a year before AI consumption, and Atomicwork’s Professional plan is reported from $25,000 a year plus $1 to $3 per outcome. Budget Serval as a six-figure platform line, and size it against what you would pay ServiceNow plus its implementation partner, not against a per-agent help desk.
Best for
IT and operations leaders at cloud-native companies of 200 to 2,000 employees, often engineering-led and already on Okta and Slack or Teams, who want to replace a helpdesk, or retire ServiceNow, and automate access requests and repeat IT, HR and finance tickets. Published customer results support the claim at that scale: SeatGeek reports going from 7,000 manual tickets to 50% automation in 60 days, Perplexity reports more than 50% of incoming requests completed automatically, and Lucid says it replaced ServiceNow in 45 days and moved 9,000 people to one front door in Teams.
It is the wrong pick for a company under about 150 employees without dedicated IT headcount, for teams that need endpoint management (Serval has no patching, monitoring or remote control and reads device state through your MDM), and for customer-facing support queues.
Watch-outs
- Agents break when systems change shape. G2 reviewers, including a design-partner customer, report that agents “can get confused” when a field is renamed or an integration changes, and that some behavior was erratic. The guard: put every published Catalyst workflow under an owner, run a weekly test request against your top 20 workflows, and keep ServiceNow or your old desk in read-only fulfillment until the automation rate holds for a full quarter.
- Opaque pricing and a vendor-staffed deployment. No public price, a paid pilot and Serval engineers building your first workflows can leave you dependent on the vendor. The guard: get the platform fee, overage unit, renewal cap and the pilot’s automation target in the order form, and require that every workflow ships as readable TypeScript your team can export.
- Short track record. The company is two years old, roughly ten times bigger than in January, and customer proof is concentrated in tech companies. The guard: ask for two references in your industry and headcount band, and confirm the SOC 2 Type II report and data-residency terms before go-live.
Alternatives, and when to pick them instead
- ServiceNow — the market-share leader. Pick it at 2,000+ employees when IT already runs its CMDB and cases cross IT, HR and customer service. Its standalone Flow service desk, launched in October 2026 at a published $2 per conversation, is the cheaper test of whether you need Serval at all.
- Jira Service Management — the other incumbent most teams price against. Pick it when engineering lives in Jira and IT and dev tickets should share one backlog, at about $20 to $51 per agent a month.
- Atomicwork — the other fast-growing AI-native service desk. Pick it when you want a published starting price and outcome-based billing instead of a six-figure platform fee.
- Ema — pick it when the goal is a general agent platform across functions rather than a system of record for service tickets.
The default for a 200 to 2,000-person company replacing a helpdesk: pilot Serval against Atomicwork on the same 30 days of real tickets, and take the one that hits your automation target at the lower 3-year cost.