What it does
6 features
ServiceNow AI Agents · Agents and orchestration on an enterprise platform
Product and access
The name ServiceNow AI Agents is current on the product page. Since May 2026, ServiceNow Otto has been the shared AI experience that routes requests to agents and combines Now Assist, Moveworks and AI Experience. The former AI Control Tower is called Super Intelligence Control Tower on the product page, and the old address redirects there; community posts sometimes still use the old name.
Core features
6 features from provider sources · checked on 5 November 2026. Availability and limits are directly below.
Agentic workflows
An agentic workflow describes the business goal, such as triaging cases. According to the product page, agents plan tasks for it, use tools and adapt their approach.
Source 1AI Agent Studio
You build and customize your own agents in natural language and set guardrails. According to the page, AI Agent Advisor suggests use cases from your data.
Source 1Action Fabric
Connects third-party agents and tools: A2A for communication between agents, an MCP client for external tools and data.
Source 1Voice, web and desktop agents
The product page names agents for phone, mobile and web, for browser-based apps and for desktop apps without an API.
Source 1Super Intelligence Control Tower
Central place to inventory, secure and measure AI agents, including third-party ones. The page names a kill switch and the AI Passport as proof of identity and permissions.
Source 2
Usable AI models
Fixed choice across providers
The ServiceNow docs list the Now LLM Service, Azure OpenAI, Google Gemini and Claude via AWS for skills and AI agents. Admins define the allowed options in AI Control Tower and set preferences per skill in the Now Assist Admin Console. Not all providers are available with in-country contracts.
Model families named
- In-house model
- OpenAI (GPT)
- Google (Gemini)
- Anthropic (Claude)
- Model choice
- You choose
- Your own API key
- Not stated
- Router or custom endpoint
- Not stated
- Local models
- Not stated
- Depends on
- Admin approval, Region
The rating covers the generally available main access. Partly documented · Sources accessed on
ServiceNow Docs: Yokohama TMT, PDF · ServiceNow: Otto press release
Does it fit my task?
2 practical tasks · 5 requirements and limits
A good fit if …
- your IT, service or HR workflows already run on ServiceNow
- you want to control agents with role masks and an approval mode per tool
- you want to inventory and secure your own and third-party AI agents in one place
Less suitable if …
- you have no ServiceNow instance and want to get started quickly on your own
- you need public prices or fixed costs per run
- you want to check data paths without access to the support portal
Editorial assessment based on provider documentation. Not a product test.
Practical tasks
Editorial examples based on documented features. No product tests performed.
Triage incidents
- Your input
- New incidents in ITSM and an agentic workflow triggered when a record is created.
- Possible result
- Agents classify incidents and prepare next steps.
- What to check
- Set high-risk tools to "Supervised" and check the trigger for recursion risk.
Build a custom agent with role masking
- Your input
- A role description in natural language and matching flow actions in AI Agent Studio.
- Possible result
- An agent that works within the masked roles of the calling user.
- What to check
- Check the licence tier, role mask and assist consumption during testing.
Have a team of agents classify incoming incidents and process them after approval.
Your possible prompt
Design a process for new requests: capture information, sort it by topic, and create a proposed response. Plan an approval step before external actions.
Adjust the task, data and permissions.
Requirements and limits
Requirement
Licence and instance
According to an employee FAQ you need Pro Plus, Enterprise Plus or an AI package such as Foundation, Advanced or Prime. Prime is said to be required to create or significantly change agents.
Source 7Limitation
Consumption during testing too
According to an employee FAQ, non-production use and tests in AI Agent Studio also consume assists.
Source 7Limitation
Agents are not deterministic
According to an employee FAQ, the same input can lead to different results. Data passed on by the model can change in the process.
Source 7Not yet confirmed
Rollout of Otto
According to the press release, Otto starts through EmployeeWorks and the Control Tower and is due to reach all products in the following year. It does not say which instance gets what and when.
Source 4Not yet confirmed
Model default
According to an employee post, since 9 July 2026 prebuilt agents without a fixed model choice move to third-party providers with the store update. The post does not say which provider applies to which agent.
Source 8
Can I use it?
Availability still unclear
Europe & regional availability
Support article KB2689562 on data residency and privacy for AI showed only a loading message when retrieved. The white paper on the ServiceNow Protected Platform for the EU could not be retrieved. Neither is therefore evaluated.
Reviewed scope: Hosted platform; European data paths for AI open
Reviewed on:
Availability still unclear
Protected Platform for the EU
An employee post says SPP EU customers use the same regional deployments as commercial EMEA customers. It does not explain how third-party providers handle data location for these customers.
Region, reason and source
Region covered: EU
Documented reason: The white paper could not be read, so conditions and scope remain unchecked.
Availability still unclear
Model choice and providers
According to an employee post, prebuilt agents move to Azure OpenAI, AWS Claude or Google Gemini. At instance level you can return to Now LLM Service in AI Admin Hub. Data residency can limit the choice.
Region, reason and source
Region covered: EU/EEA, United Kingdom and Switzerland
Documented reason: Depends on configuration and provider; the post does not give the regions per provider.
Availability still unclear
EU AI Act in the Control Tower
According to its page, the Control Tower includes content for the EU AI Act. This says nothing about the availability or data location of agents in Europe.
Region, reason and source
Region covered: EU
Documented reason: This is governance content, not a regional statement.
Accounts & permissions
Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use.
Configuration covered: Vendor platform with role control in the instance
Sources reviewed:
Show all 3 points
Documented
Data access & team sharing
According to the post, role configuration and role masking are mandatory and limit workflow, agent and tool in sequence. Tools are checked against the run-as identity.
Evidence
Memory and conversation continuity
2 of 6 criteria documented
Short-term memory applies per conversation; long-term memory is stored in sn_aia_memory and controlled through properties; official documentation could not be read.
Short-term and long-term memory according to employee posts
Sources accessed:
Documentation by criterion
StorageDocumented
Storage and scope
Short-term memory applies to one conversation and is shared between agents. Long-term memory persists across sessions in the sn_aia_memory table, in vector format.
SelectionDocumented
Selection and timing
Properties turn on long-term memory. sn_aia_ltm_category defines which kinds of information may be stored, such as device preferences.
Edit & deletePartly documented
Inspect, edit and delete
Admins and developers review and configure categories. The guide names no interface where end users view or delete their own entries.
Access boundariesPartly documented
Agent and channel boundaries
sn_aia_ltm_category_mapping controls which agents access which categories. The guide does not describe boundaries between channels.
RetrievalPartly documented
Retrieval
The guide names the vector format and the property for use in the agent. It does not say how entries are selected or weighted.
New conversationsPartly documented
Session reset and compaction
According to the FAQ, context is passed on in short-term memory, and the context window is 128K tokens; exceeding it can cause unpredictable behaviour. The FAQ does not mention reset or compaction.
Cost and effort
3 plans · Prices as of 5 November 2026
Prices and included usage
Pricing checked
No public prices. The ITSM pricing page shows Foundation, Advanced and Prime and only offers "Get Custom Quote".
| Plan | Price | Included |
|---|---|---|
| ITSM Foundation | No public price; custom quote | According to the page, agents and skills for task-based support. |
| ITSM Advanced | No public price; custom quote | According to the page, agentic workflows and AI Voice Agents. |
| ITSM Prime | No public price; custom quote | According to the page, AI Specialists and AI Agents for ITSM. |
- Packages of other products were not reviewed. An employee guide names 25 assists for agentic workflows with 0 to 4 tools, 50 with 5 to 8 and 150 with 9 to 20. This is no price list, and it gives no assist prices.
- Third-party pricing was not used as evidence.
Setup
Requires a ServiceNow instance and a matching licence. You activate prebuilt agents and build your own in AI Agent Studio using natural language. The entry point is a demo or quote.
Operation & oversight
Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured.
Deployment covered: Operation in the ServiceNow instance; models from ServiceNow or third parties depending on configuration
Sources reviewed:
Editorial assessment
Permissions, approvals and logs are described in detail, but often in community articles by employees. You need to check resuming, retention periods and data paths in the documentation, support articles and contract.
Show all 6 points
Documented with conditions
Files and working state
According to an employee FAQ, the execution plan and tasks are stored in instance tables (sn_aia_execution_plan, sn_aia_execution_task). The reviewed pages give no retention periods.
EvidenceDocumented with conditions
Sign-ins
The agent runs as a dynamic user or AI user. The reviewed pages do not describe how external tools manage credentials or which SSO applies.
EvidenceDocumented
Starting further runs
The employee guide names triggers on record creation or update, on a schedule and on inbound email, plus manual start in the Otto panel. It warns about recursion and names limits.
Evidence
Evidence and open questions
9 sources · 5 open points
Open points
The reviewed sources contain no reliable information on these points. A gap says nothing about the product.
Europe: Availability still unclear
Requirements and limits: not yet confirmed
Sources and evidence
Product pages, the ITSM pricing page and the Otto press release reviewed. Technical details come from community articles marked "ServiceNow Employee"; they are not official documentation, and one author explicitly calls his governance recommendations his own. The support article on data residency (KB2689562) and the SPP EU white paper could not be read. Example tasks are editorial. Hands-on tests are still pending.
Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.
Provider
ServiceNow offers a platform for enterprise workflows with integrated AI agents.
- Operator
- ServiceNow
- Organization
- Provider
- Website
- www.servicenow.com
Changes to this profile 1
AI agents on the ServiceNow AI Platform that work in teams through an orchestrator and agentic workflows to run tasks in IT, customer service and HR.
These notes describe editorial changes. Source and pricing review dates appear in their respective sections.