ServiceNow
Also known as: Now Assist, ServiceNow AI Platform, ServiceNow Otto
Workflow platform whose AI agents, built in AI Agent Studio, run with supervised or autonomous tools, long-term memory, pre-release evaluations and browser actions, callable by any outside agent through Action Fabric.
ServiceNow builds AI agents into the workflow platform that runs IT, HR, customer service and security operations for large enterprises. Customers deploy out-of-the-box agents or build their own in AI Agent Studio, and the Autonomous Workforce adds AI specialists that take on whole jobs, orchestrating teams of agents from start to finish.
Agents start on their own when a configured condition is met, and also work conversationally through the ServiceNow Otto panel. They act through tools added in AI Agent Studio, reach other systems over MCP and A2A, and can operate external websites directly. With AI Desktop Actions an agent opens the site in its own browser tab, carries out the steps and reports each action back. Long-term memory stores facts about each user in categories mapped to specific agents, and Knowledge Graphs give agents the relationships in the customer's data.
Oversight is set per tool. Any tool can run in supervised mode, so a person oversees the action before it executes, while lower risk tools run autonomously. Agents are tested before release in agentic evaluations that score runs against defined datasets, sessions and traces are kept for analysis with retention the customer controls, and administrators choose the model provider from Now LLM Service, AWS Claude, Azure OpenAI and Google Gemini.
From outside, Action Fabric lets Claude, ChatGPT, Gemini or any other agent run governed ServiceNow actions over MCP and A2A. Security rests on SAML single sign-on, role-based multi-factor authentication and access control, and a separate Government Community Cloud serves FedRAMP customers. Since April 2026 AI is included in every ServiceNow product, in tiers that run from AI assistance to fully autonomous operations, with prices quoted through enterprise sales. AI Passport, a runtime credential that limits what each agent may do, with graduated responses up to a kill switch, is expected in the fourth quarter of 2026.
Vendor details
Canonical URL
https://www.servicenow.com/products/ai-agents.html
Category
Enterprise operations agent
Subcategory
Platform — AI Agents, AI Agent Studio, Autonomous Workforce and Action Fabric
Funding status
Public company listed on the New York Stock Exchange under the ticker NOW. ServiceNow Otto brings Now Assist, Moveworks and AI Experience together as one assistant across ServiceNow products.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
AI agents use tools added in AI Agent Studio, such as catalog items and Knowledge Graphs, and reach outside systems as an MCP client. Action Fabric exposes ServiceNow to external agents, so any agent, including Claude, ChatGPT and Gemini, can run governed ServiceNow actions headlessly over MCP, A2A and traditional protocols, with an MCP Server Console controlling what is exposed. AI Desktop Actions let agents work on external websites through a Chrome extension.
In practice
Your service desk is flooded with routine requests. Out-of-the-box ServiceNow agents take them on, starting automatically when a request meets their trigger, with sensitive tools set to supervised mode so a person oversees those actions.
A legacy web application has no API. An AI agent with desktop actions opens it in a browser tab, works through the screens, and reports each step back to the person who asked.
Your teams already use Claude or ChatGPT. Through Action Fabric those assistants run governed ServiceNow workflows over MCP, with an MCP Server Console deciding exactly what is exposed.
Sources & related URLs
Research sources
Agentic Index coverage score
13.5 / 14 capabilities · 96%
| Integrations & Tool Calling | Full |
|---|---|
|
Agents act through tools added in the guided setup, such as catalog items, and reach outside systems as an MCP client through Action Fabric, which ServiceNow describes as MCP server, MCP client and A2A together, with A2A used to orchestrate multiple systems and processes. Sourceservicenow.com/platform/action-fabric.htmlread 2026-09-18 |
|
| Workflow Orchestration | Full |
|
The Autonomous Workforce deploys AI specialists that do whole jobs and orchestrate teams of AI agents to deliver outcomes from start to finish. Action Fabric uses the A2A protocol for complex multi agent workflows across systems and processes. Sourceservicenow.com/products/ai-agents.htmlread 2026-09-18 |
|
| Knowledge Grounding & RAG | Full |
|
When an AI agent is set up, tools and the customer's own information are added to it, including Knowledge Graphs that give the agent the relationships between real world entities to improve its outputs, for example a Knowledge Graph added to an approval agent. New knowledge is added this way, without retraining the model. Sourceraw.githubusercontent.com/servicenow/servicenowdocs/australia/markdown/intelligent-experiences/add-tool-aia.mdread 2026-09-18 |
|
| Human Oversight & Guardrails | Full |
|
Each tool an AI agent uses can be set to run in supervised execution mode, chosen as the execution mode when the tool is added in the guided setup, so a person oversees the tool's actions and the potential impact of an agent not executing as expected is contained, while other tools run autonomously. The mode is set per tool, so low risk and high risk actions can run under different levels of autonomy, and AI Agent Studio also sets guardrails on each agent. Sourceraw.githubusercontent.com/servicenow/servicenowdocs/australia/markdown/intelligent-experiences/add-tool-aia.mdread 2026-09-18 |
|
| Security, Identity & Governance | Full |
|
ServiceNow's Trust and Compliance Center lists FedRAMP authorization for its Government Community Cloud. Platform controls include SAML 2.0 single sign on, including SSO across multiple providers, role based multi factor authentication and access control, and agents run as dedicated service accounts with scoped roles. Sourceservicenow.com/company/trust.htmlread 2026-09-18 |
|
| Observability & Auditability | Full |
|
Agent runs are recorded as sessions, traces and spans kept for analysis, and the customer sets the retention. Session, trace and span data is kept for 30 days by default, and a system administrator can shorten the window or stop retaining trace data for data minimization requirements, and choose whether the data stays available for analysis. AI agents using desktop actions also report each action they take back to the user in the ServiceNow Otto panel. Sourceraw.githubusercontent.com/servicenow/servicenowdocs/australia/markdown/intelligent-experiences/mon-ai-configure-trace-retention.mdread 2026-09-18 |
|
| Memory & State Persistence | Partial |
|
AI agents keep a long term memory of user specific facts and preferences. When an agent runs, an LLM checks the interaction for facts that match the agent's configured memory categories and stores them as semantic memory records in the AI Agent Memories table, scoped to the user and category and kept apart from business data. Categories such as software preferences, workplace context and communication style are defined globally and mapped to individual agents, and agents can also learn from the outcomes of past executions. Administrators decide whether agents may retain user information, apply it in conversations or learn from past runs, and the stored facts can be viewed by user and agent. How long memories are kept, and how one user's memories are deleted, is not stated. Sourceservicenow.com/docs long term memory categories and set up long term memoryread 2026-10-25 |
|
| Deployment & Data Residency | Full |
|
ServiceNow offers a separate environment for regulated customers, the Government Community Cloud with FedRAMP, alongside its standard cloud, and customers that need a government authorized boundary run there. Sourceservicenow.com/company/trust.htmlread 2026-09-18 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Buyers deploy out of the box AI agents or build their own, and ServiceNow names ready made agents across domains, such as change request planning, change template suggestion, supplier approval, asset evaluation and post incident review agents. Sourceservicenow.com/products/ai-agents.htmlread 2026-09-18 |
|
| Triggers & Channel Coverage | Full |
|
In the guided setup, an administrator adds triggers that run the AI agent automatically when certain conditions are met, so the agent starts when an event occurs rather than only in chat. Agents also work conversationally through the ServiceNow Otto panel and Virtual Agent. Sourceraw.githubusercontent.com/servicenow/servicenowdocs/australia/markdown/intelligent-experiences/add-trigger-aia.mdread 2026-09-18 |
|
| Model Flexibility & Routing | Full |
|
An administrator chooses the model provider for AI agents. In AI Agent Studio settings, the model provider tab leads to AI Admin Hub, where the default provider is selected from Now LLM Service, AWS Claude, Azure OpenAI and Google Gemini, and models can be tested in the skill kit before switching. ServiceNow notes that agents should be retested after the global model changes, and that in some regions using a different provider requires consent. Sourceraw.githubusercontent.com/servicenow/servicenowdocs/australia/markdown/intelligent-experiences/select-aia-llm.mdread 2026-09-18 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Action Fabric lets any AI agent, including Claude, ChatGPT and Gemini, execute governed ServiceNow actions headlessly through MCP, A2A and traditional protocols. The MCP Server Console is a central control plane for exposing ServiceNow resources to external AI agents with native governance, and ServiceNow can also act as a secondary agent over A2A. Sourceservicenow.com/platform/action-fabric.htmlread 2026-09-18 |
|
| Testing, Debugging & Optimization | Full |
|
Agentic evaluations run AI agents and agentic workflows against defined datasets to verify quality before deployment. Each run scores performance, identifies issues and suggests optimizations, with custom metrics added through a parser tool and a results page for analysis. Sourceraw.githubusercontent.com/servicenow/servicenowdocs/australia/markdown/intelligent-experiences/agentic-evals.mdread 2026-09-18 |
|
| Browser & Computer Use | Full |
|
With AI Desktop Actions, an agent triggered from the ServiceNow Otto panel performs tasks on an external website or web application, opening the site in a separate browser tab, carrying out the steps and reporting its actions back in the panel, through the ServiceNow Web Automation Chrome extension connected to the instance. Adaptive desktop actions handle interactions beyond custom UI, including triple-click, double-click, right-click, native prompts, dialogues and alerts, administrators configure the allowed websites, and the agent runs as a dedicated service account. Sourceraw.githubusercontent.com/servicenow/servicenowdocs/australia/markdown/intelligent-experiences/na-ai-wa-access-using-nap.mdread 2026-09-18 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
ServiceNow's August and September updates to AI Control Tower add an AI Asset Inventory with domain separation, integrations for governance and monitoring, and expanded oversight features.
Bears on: Observability / auditability
View sourceServiceNow introduced AI Specialists, a new agentic capability designed to autonomously handle end-to-end workflows such as incident resolution. The first generally available specialist, the L1 Service Desk AI Specialist, can triage incidents, diagnose issues, apply fixes, and communicate with requesters. It operates in either an autonomous Autopilot mode or a human-in-the-loop Copilot mode.
Bears on: Workflow orchestration
View sourceServiceNow rebranded Now Assist for Creator to ServiceNow Otto for Creator and introduced several updates to its Build Agent. The release adds the ability to define custom skills and rules for the agent, allows developers to run and roll back scripts directly from the chat panel, and enables the agent to build within isolated Developer Sandboxes.
Bears on: Human approval / guardrails
View sourcePricing
No public price. AI is included in every ServiceNow product, in tiers quoted through enterprise sales.
Product and package subscriptions sold through enterprise sales, with AI included in each product and tiered from AI assistance to fully autonomous operations.
What is public
No public rate. ServiceNow publishes no list pricing, only that AI is included in every product and offered in tiers. Platform, product and tier pricing are negotiated through enterprise sales and partners.
Billing mechanics
Product and package subscriptions sold through enterprise sales. Since April 2026 each product includes AI, data connectivity, workflow execution, security and governance, in tiers from AI assistance to fully autonomous operations.
Cost watchouts
AI now comes inside each product's tier, not as a separate purchase, so cost follows the tier and the products bought. ServiceNow does not publish the tier prices, what each tier allows or how heavy agent use is metered.
Variable cost rationale
AI is included in every ServiceNow product, in tiers that run from AI assistance to fully autonomous operations, so the tier chosen and the products bought set most of the bill. ServiceNow publishes no tier price, usage allowance, overage rate or cap, and has not said publicly how agent usage is metered under the tiers it introduced in April 2026.
Additional watchouts
Moving up the tiers, from AI assistance to agentic automation to fully autonomous operations, is where cost grows, and ServiceNow publishes neither the tier prices nor their limits.
Sales call required
Yes, required for paid access
Free / trial
No self serve tier. Demos and trials run through sales and partners.
Lowest paid plan
Not published
Key ambiguities
No public rate exists for the platform, products or AI tiers. All pricing is quoted through sales and varies by product mix and negotiated terms.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Abacum — FP&A platform for growing companies whose AI Modeler builds, edits…
- Adopt AI — AI native accounting and bookkeeping firm whose agents run…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
Alternatives to ServiceNow
The closest documented capability profiles to ServiceNow among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Appian13.5 / 14Matches ServiceNow across all 14 documented capabilities
- Pega13.5 / 14Matches ServiceNow across all 14 documented capabilities
- UiPath13.5 / 14Matches ServiceNow across all 14 documented capabilitiesServiceNow vs UiPath →
- Microsoft14.0 / 14Fuller documented coverage on Memory & State PersistenceServiceNow vs Microsoft →
- Tungsten Automation13.0 / 14A lighter documented profile than ServiceNow
- Writer13.0 / 14A lighter documented profile than ServiceNow
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded