Serval has closed a $47 million Series A funding round to put agentic artificial intelligence at the heart of IT service management. The round is being led by Redpoint Ventures, with First Round, General Catalyst, and Box Group also participating in the raise — a testament in itself to the attention that it appears to be grabbing early on here; it comes alongside an impressive list of early customers: companies like Perplexity, Mercor, and Together AI. The pitch is straightforward and timely: it’s time to move past chatbots and make tickets into deterministic, policy-based automations that actually do work.
Instead of a single monolithic helper, Serval is divided between two agents. One is an in-house developer who creates and maintains the automations that manage daily tasks such as software authorization, device authorization, or access requests. One is the help desk front end that interprets user intent and makes calls to those tools under very specific conditions. It is a design intended to achieve the flexibility of large language models while retaining explicit control over execution.

Why ITSM workflows are poised for agentic AI adoption
IT service desks are still inundated with repetitive, policy-enforced requests, despite organizations all over the world standardizing on frameworks like ITIL. Popular metrics from MetricNet and HDI have long demonstrated that Level 1 tickets come with significant per-incident costs, and high-volume/low-complexity work drags down teams. And ServiceNow’s multibillion-dollar business points to how big the operational demand is within large enterprises.
Service operations are already being transformed by generative AI. A majority of enterprises will add intelligence into service workflows, real-time interaction channels, and self-service, according to analysts with Gartner as well as Forrester, in the next few years. In general, the goal is faster issue resolution and fewer escalations. The next step is to use that linguistic comprehension to drive controlled action — and quantify the improvements in first-contact resolution and mean time to resolution.
Two coordinated agents with guardrails for control
Serval’s “builder” agent authors the underlying automations and policies, and then abstracts them out as deterministic tools. Managers approve and configure the permissions of each tool: which users may invoke it, what sort of authentication is required, the hours that it can be run, and the systems that it is permitted to touch. Since the tools are software, they could include such complex checks as multi-factor authentication status, device compliance posture, or time-bound change windows.
The orchestrator: the “help desk” agent. It understands a request, chooses the right tool for it, and executes only within the guardrails that managers put in place. That divide allows teams to track who made what, when it changed, and how it is used. And if policies change, the builder agent can modify the codebase rather than teaching staff to retrain a general-purpose bot.
Take a common situation: a contractor requires access to a private Git repo. The help desk agent can confirm identity, verify contract status against HRIS, check device compliance via a (potential) endpoint management tool, enforce MFA, and also issue a scoped repo role — each action being audited and automatically scheduled to be revoked in the future. If any check doesn’t pass, it gets kicked up to a human who has all the context.
The company’s management has stressed that the priority is visibility and control: automation is a secondary objective. Enterprise consumers who are a little squeamish about random outputs from some deep learning model or other would prefer to have language used for generating and routing work, but want changes made by something with predictability and auditable facts.
Early customers and the anticipated impact on ITSM
Early users such as Mercor, Perplexity, and Together AI show that Serval’s tools are hitting a chord with engineering-heavy organizations where speed is lucrative but regulatory non-compliance simply isn’t an option. Performance metrics are not revealed, but mature programs across the ITSM market often focus on 30–50% automation of repetitive Level 1 tickets and substantial cycle time reductions for access and provisioning requests.

Leaders will monitor measurable results, some of which are:
- First-contact resolution
- Mean time to resolution
- Backlog burn-down
- Change failure rates
Equally critical are governance indicators:
- Policy conformance
- Audit completeness
- Ability to prove least-privilege enforcement to security teams or regulators
Competition in ITSM and key questions for buyers
Serval is entering an increasingly crowded field. Incumbents such as ServiceNow, Atlassian, and Zendesk are building generative capabilities into their platforms, while RPA vendors including UiPath and Automation Anywhere are tackling next-door automation needs. Serval’s differentiator lies in the clear separation between tool generation and tool execution, stressing determinism, permissions, and audit throughout the entire lifecycle.
Potential buyers will be asking questions in several categories:
- Integration depth across identity and device management, directory services, ticketing, and collaboration tools
- Approval-flow strength with human-in-the-loop checks
- Logging for SIEM compatibility and compliance needs
- Deployment model options to meet data residency and security standards
By no means are certifications such as SOC 2 or ISO 27001 special to the cloud database space; rather, they represent a bar for entry within this group.
What comes next for Serval and enterprise ITSM
With new cash and credible customers, Serval’s next task is scale and reliability: growing integrations, hardening guardrails, and demonstrating high success rates in the messy edge cases that characterize enterprise IT.
Anticipate further investment in retrieval-augmented generation, function calling, and policy engines to ensure that language models stay expressive while the upstream actions stay predictable and reversible.
The broader trend is clear. Service desks are advancing from conversational chatbots that “answer questions” to agentive systems that can be directed to take action under controlled guidance. If Serval can keep turning tickets into auditable, policy-friendly changes as a matter of routine, and without taking anyone by surprise (such that the alignment sticks: if indeed it does), then it has a shot at becoming an IT operations “DB for AreacodeNEXT”.