Strengths
Cautions
Learn More
1. Expert Insights Verdict
Serval is an AI-native platform combining service management, access management, and an asset database. It’s aimed at enterprise IT teams, and works either as a full ITSM replacement for ServiceNow, Jira Service Management, or Freshservice, or as an automation layer sitting above them with two-way sync.
Serval provides multiple agents in one platform. First, a help desk agent that can resolve employee requests in Slack, Microsoft Teams, email, voice, or a web portal. A second agent called Catalyst is the highlight for admins. It builds automations, reports, and workflows based on natural language prompts and integrates across any 3rd party application via API. Automations are generated as readable TypeScript rather than prompt chains. This means they are deterministic, testable, auditable, and editable by an administrator who wants to check what the agent is building.
Serval’s access management component covers just-in-time provisioning, scheduled user access reviews, and identity governance. Serval Databases covers hardware and software asset management and a CMDB, building a context graph of devices, licenses, users, and configuration items ingested from MDM, identity providers, HR systems, and procurement tools. Both are native to the platform, which means an access request or an asset question is handled by the same agent and against the same data as everything else.
We think Serval is a strong fit for enterprise IT teams with high ticket volumes looking to build AI-powered helpdesk agents and more streamlined workflows.
2. Fast Facts
- Headquarters: San Francisco, CA
- Founded: 2024
- Ownership: Private, venture-backed
- Funding: A $75m Series B in December 2025 led by Sequoia Capital valued the company at $1bn, following a $47m Series A in October 2025 led by Redpoint Ventures.
- Security and compliance: SOC 2 Type II. Serval also states adherence to HIPAA and GDPR, and publishes a trust center
- Deployment: Cloud, hybrid, and self-hosted
3. Serval’s Approach
Serval defines itself as an AI-native IT service management (ITSM) solution, competing in a new generation of startup providers that is built with generative AI at the core of the solution. Serval can also be deployed as an add-on layer to existing ITSM deployments if you do prefer to stick with your existing ITSM provider.
Serval argues there are three weaknesses with traditional ITSM approaches. First, admins have to configure every new application and team by hand. Second, automation is difficult to scale so engineers keep resetting passwords and provisioning access manually. And third integrations can break whenever a vendor changes an API.
Serval’s answer has been to build AI agents directly into its ITSM platform. The Help Desk Agent is what employees interact with. It can resolve common requests as vague as “my laptop is slow” or “I want to be an admin in the design tool.” When it acts, it calls a published automation with fixed inputs, fixed outputs, and full logging. It behaves the same way every time.
Catalyst is the more interesting part of the platform for admins. It’s a GenAI engine that builds reports, workflows, dashboards and automations for admins based on natural language prompts. You can simply describe what you want to build and Catalyst writes the underlying TypeScript, generates the form, sets the trigger, and stages it as a draft.
It can also work from evidence in place of instruction. If you provide historical ticket data in ServiceNow or Jira for example, it will analyze and draft automations for the highest-volume problems it finds.
Serval also handles access requests and treats them as provisioning actions rather than as tickets to be fulfilled by hand. Just-in-time grants are time-bound with automated removal, which reduces standing privilege, and provisioning runs through IdP group membership, direct API calls, custom workflows, or a manual task where no automation path exists.. Access profiles restrict who can request what before a request is made. Scheduled user access reviews produce the evidence SOC 2 and ISO 27001 require.
Since early 2025, buyers have increasingly focused on the accuracy and completeness of configuration and knowledge data, because conflicting or outdated records are a primary cause of inaccurate AI output. Serval Databases reconciles conflicting asset records from across connected systems into a single schema, with an admin-set source-of-truth priority for each field, which addresses that dependency directly.
Serval has also recently launched Proactive Agents: scheduled Catalyst sessions that run against a team’s tickets, workflows, databases and connected apps, investigating issues, diagnosing problems, and drafting fixes for admin review. Each agent runs as a designated user and is bounded by that user’s permissions.
The company ships a library of prebuilt agents covering ticket digests, licence reclamation, service-desk quality review, ticket trend analysis, automation-opportunity discovery, and device remediation, and admins can build their own for anything else, from monitoring AI spend against a budget to sweeping for unused accounts.
4. Market Position
Serval is competing in a very established market that is changing quickly with the adoption of generative AI. There has also been a lot of consolidation – from November 2025 to early 2026, ServiceNow acquired Moveworks, Automation Anywhere acquired Aisera, and Zendesk acquired Forethought.
Broadly three groups of vendors compete for AI service management budgets. First, ITSM platforms including ServiceNow, Jira Service Management and Freshservice have added AI to their ITSM solutions that have been established for many years.
Second, a set of AI-first vendors have built conversational agents and automation layers that run on top of whichever platform a customer already owns. Finally, a new generation of start-up ITSM providers have launched AI-native ITSM solutions fully built around GenAI capabilities.
Serval competes across these areas. It sells a full system of record covering request, incident, problem and change management, a CMDB, access management and a service catalog, which puts it against the established platforms.
It also sells AI agents which can build the automation workflows, which puts it in the group of the AI-first vendors. The two-way sync with ServiceNow and Jira Service Management lets a customer buy it as either one: an automation layer or a complete ITSM platform.
The target buyer is an IT team whose ticket volume is growing faster than its headcount and whose automation coverage has stalled. Serval’s customers include Spotify, Fox Corporation and Snowflake alongside Ramp, Perplexity, Together.ai, Mercor and Cribl, and the company reports an average deployment spanning 13 departments.
5. Key Features and Use Cases
Resolving employee requests wherever they arrive
Serval’s AI helpdesk agent can be deployed in Slack, Microsoft Teams, email, voice, and a web portal, and can be embedded in a ServiceNow portal or white-labeled. Requests can arrive in a public channel or a direct message.
It responds directly to user queries with AI-generated answers, even to very vague prompts. Answers are drawn from connected sources including Notion, Confluence, Google Drive, SharePoint, ServiceNow and Freshservice, or from Serval’s own knowledge base, with links back to the source.
Serval treats this as table stakes and argues the interesting work starts when the agent takes action. Asked about a slow laptop, for example, Serval’s agent can look up the device in the connected MDM, check eligibility for replacement, and either resolve or escalate the issue directly.
Where a full resolution is possible it runs end to end: provisioning a corporate card, filing a time-off request in Workday, or granting application access. Requests can be surfaced as forms.
Every interaction produces a ticket even when no human touches it, which preserves the audit record.
Building automations in natural language
Catalyst is the automation builder for admins. A prompt as short as “build me an MFA reset workflow for Okta” produces a working automation with its trigger, steps, and permission model laid out, plus the generated code beneath it. Serval’s position is that the code should be visible on principle: administrators need to be able to read exactly what the AI wrote.
The same interface can build forms, dashboards, access policies, skills, asset schemas, and journeys, the multi-step onboarding and offboarding processes that coordinate tasks across teams.
Catalyst has API access to Serval itself and to any connected application, so a request to analyze ticket and access data and produce a dashboard results in the underlying fetch workflows being written and the dashboard assembled from them.
It also works from your history. If you provide access to ServiceNow or Jira ticket data, Catalyst will analyze volume, report what it finds, and draft automations for the highest-value unautomated problems, skipping anything already covered.
Identifying what to automate next
Serval identifies its own automation gaps in two ways. It watches ticket patterns continuously and, when it spots a request that was resolved by hand but could have been automated, raises a suggestion: a new workflow, a new or updated skill, an app connection, or a knowledge article. Suggestions surface in the team’s Skills and Workflows views, where an administrator can accept, edit and accept, or dismiss them, and every decision is retained in a history for audit.
Separately, a weekly agent reviews recent tickets for repetitive manual requests with no coverage and drafts the corresponding workflows through Catalyst, so nothing it proposes reaches production without an admin reading the generated code and releasing it.
Governing what the agents can do
Access profiles control which users can trigger automation, scoped by team, group, or user attribute. Approval requirements are set per automation and support multi-step chains that approve in sequence, group-based approvals, automatic routing to the requester’s manager, and custom rules that can auto-approve, deny or escalate on the team’s own logic. A workflow can also pause partway through to collect an approval and resume once it is granted.
Automations are routed through staging and review before reaching production in the same way code would be.
Managing access and running reviews
Access requests are one of the largest categories in most help desks. Serval supports just-in-time provisioning and automated deprovisioning, reducing standing privilege. Provisioning runs through IdP group membership, direct API calls, custom workflows, or a manual task, and access profiles restrict who can request what in the first place.
Scheduled user access reviews support SOC 2 and ISO 27001 evidence requirements. For AI tooling specifically, Serval points to budget monitoring alongside provisioning, which is becoming its own category of request.
Reconciling asset and configuration data
Serval Databases covers hardware, software licences, SaaS subscriptions, and configuration items against custom schemas, with configurable fields, statuses, and relationships between assets.
It ingests from MDM, HRIS, identity providers, procurement tools, and custom APIs, suggests how each external field maps into the schema for an admin to accept or adjust, and deduplicates on a designated key field across overlapping systems. Where two sources feed the same field, the admin sets which one takes priority.
Every value carries the identity of the source that wrote it, with full field-level history, so the reconciled record stays traceable to the systems it came from rather than becoming an unexplained parallel inventory.
Assets link automatically to the tickets that involve them, and the platform tracks ownership history, lifecycle, and usage for compliance review, with end-of-life tracking based on manufacturer end-of-sale and end-of-support dates.
Finding problems before they are reported
Proactive agents investigate on a set schedule and draft fixes for review. Serval ships a library covering service desk analysis, daily ticket digests, proactive remediation, license reclamation, and administrators can build custom agents that take action on a recurring schedule.
6. User Experience
For employees there is almost nothing to learn. A request in Slack or Teams gets a response in the same thread, and where a form is more appropriate than a conversation the agent presents one. Screenshots can be dropped into chat for visual troubleshooting.
For administrators the platform presents as a full ITSM. Tickets support request, incident, problem, and change management with custom queues and views, SLAs, due dates, categories, and custom fields, and service catalogs can be ingested from ServiceNow, Freshservice, and Jira.
Catalyst itself is a chat interface with suggested starting points across IT, HR, finance, and legal use cases, plus a quick start for administrators who do not yet know what they want to build. Generated automations appear beside the conversation as they are built, and can be refined either conversationally or by editing the fields and code directly. The interface is clean, and easy to use.
Alongside the conversation, an administrator can see which workflows ran, the individual API calls within them, the knowledge articles consulted, and the assets involved, each openable in the asset database.
From there they can respond to the employee, invoke Catalyst to take an action no published automation covers, ask the system to suggest an automation so the request does not recur, or troubleshoot what went wrong.
For deployment, Serval runs a four-phase guided pilot with a dedicated deployment engineer: connecting integrations and introducing foundational automations, identifying and building the highest-value workflows, trialing with end users and iterating, then expanding into cross-company use cases and auditing the configuration.
7. Pricing
Serval does not publish pricing. The model is a flat platform fee rather than per agent or per seat, which the company positions as protection against the runaway costs of consumption-based AI pricing. Pricing is established through a pilot that establishes usage and utilization, then fixed for the contract term.
The AWS Marketplace listing for the Serval Enterprise Platform is $500,000 for a twelve-month contract, negotiated through private offers. That is enterprise pricing, and smaller IT teams should assume the platform is not aimed at them.
Serval describes its own cost structure as scaling with usage while presenting the customer-facing fee as flat, so it is worth establishing at the outset what happens to the fee at renewal if volume grows substantially.
8. Who It’s For
Serval is built for enterprise IT teams with high ticket volumes, who need to build effective automation workflows at scale. If you want to move away from a legacy ITSM or cannot get automation coverage out of it fast enough, Serval can help address those challenges with an AI native solution.
If your problem is that ticket volume is growing faster than headcount, and the bottleneck is how long it takes to build and maintain automations, Serval addresses that directly. Catalyst generates automations from natural language and from historical ticket data.
If your security team is the obstacle to adopting AI in the service desk, Serval stands out for its governance model. The separation between the user-facing agent and the automations it can call, combined with permission enforcement at the automation level and a staged publishing flow, is a more defensible design than a single agent holding API credentials.
If you are consolidating service delivery beyond IT, the same platform extends to HR, finance, and legal, and Serval reports customers running it across multiple departments.
