By Tsuky 
An on-call engineer investigating a service outage may need information from monitoring platforms, previous incident reports, change records and asset management systems. Finding those details is only the beginning. Someone must still determine the next action and update the relevant operational records.
OpenAI and ServiceNow are working to combine these steps inside ChatGPT. Their expanded agreement introduces a plugin that can connect the assistant to ServiceNow workflows while preserving the enterprise platform's established access controls.
The announcement builds on a strategic partnership established earlier in 2026. It includes one integration that is available now, enhanced experiences in preview and several capabilities scheduled for future release.
Understanding these different release stages is essential. The companies have not launched every feature described in their October announcement.
The ServiceNow MCP plugin is available now
The ServiceNow MCP plugin is available through the OpenAI marketplace. Organizations can connect ChatGPT to their own ServiceNow instances and determine which operations users are allowed to access.
An IT employee could ask ChatGPT to find similar incidents, review information relevant to a proposed change or update an applicable ServiceNow record.
The key difference from an ordinary chatbot is that the integration can move beyond presenting information and help perform authorized operations in the underlying business system.
ServiceNow Action Fabric provides the connection to existing workflows, business rules and permissions. ServiceNow remains the system responsible for executing and governing the operations.
That distinction helps preserve existing organizational controls. Connecting an AI interface should not give someone broader access than their account already permits.
Each company must configure the connection and decide which capabilities to make available. The plugin's presence in the marketplace does not automatically expose private corporate records to every ChatGPT user.
Interactive incident handling is being previewed
ServiceNow is also developing richer interactive experiences inside ChatGPT. One example involves an on-call engineer responding to a technical incident.
The proposed interface would bring together linked monitoring alerts, recent changes and affected services. From there, an engineer could investigate the issue, assign an appropriate team and work toward resolving the incident.
The workflow aims to reduce the need to navigate between independent applications during an operational problem.
However, these richer interactive experiences are still being previewed. They should not be confused with the features already delivered through the current plugin.
The companies expect enhanced plugin capabilities by the end of 2026, without guaranteeing that every demonstrated experience will be universally accessible at that point.
The quality of an incident investigation also depends on the accuracy and completeness of the available context. An AI system may identify a plausible connection between events without establishing that they share a common cause.
Operational safeguards, approvals and the ability to reverse changes will remain necessary for sensitive production systems.
Codex is coming to AI Workflow Factory
OpenAI and ServiceNow plan to integrate Codex into ServiceNow's AI Workflow Factory during 2027.
The intention is to help developers create and modernize applications while using the business context already represented in ServiceNow.
OpenAI models are also expected to support Build Agent and Autonomous Engineer, components involved in application development and modernization.
These capabilities address an important limitation of general-purpose coding assistants. Generating functional code is not enough when an application must conform to existing business processes, access restrictions and integrations with legacy systems.
An agent working inside that context could potentially produce software better aligned with the organization's operational requirements.
This remains a planned capability rather than an established improvement. The companies have not published independent performance results showing how reliably the integration handles complex enterprise software migrations.
There is also no precise commercial release date beyond the expected 2027 availability.
Cybersecurity combines technical findings with business context
The partnership includes a planned security solution combining OpenAI's cybersecurity capabilities with ServiceNow's information about enterprise assets, identities, permissions and applications.
One goal is to improve vulnerability prioritization. Security scanners can produce substantial lists of findings, but technical severity alone does not always reveal which issue creates the most immediate business risk.
A vulnerability affecting an internet-facing application with access to sensitive information may require a different response from a similar weakness in an isolated internal environment.
By combining technical analysis with enterprise context, the partners aim to identify potential attack paths, assess which problems matter most and coordinate remediation work.
The planned workflow spans vulnerability identification, validation, prioritization and mitigation across software, workloads, networks and identities.
ServiceNow describes the effort as part of its broader strategy to move security activities toward earlier and more proactive intervention.
The product is not generally available today. Early access is expected in January 2027, with general availability planned for the second quarter of 2027.
Real-world reliability will depend on the quality of asset inventories, identity records and application relationships. Incorrect or incomplete enterprise context can distort prioritization just as easily as an inaccurate model assessment.
Voice interactions will support AI risk assessments
Another planned feature uses OpenAI voice technology inside ServiceNow's Super Intelligence Control Tower.
The proposed interaction allows a risk professional to describe an AI system or use case aloud. The assistant would populate assessment fields based on the description and ask follow-up questions when required information is missing.
The system is also intended to produce an initial risk classification from the available answers.
This could reduce the administrative effort involved in preparing assessments, but it does not eliminate the need for validation. A spoken description may leave out important details about data access, permissions or operational consequences.
The voice feature is expected to become available in the first quarter of 2027. It remains a future capability rather than a broadly deployed product at the time of the announcement.
Both companies are expanding their internal use of the partnership
ServiceNow plans to make ChatGPT Work and Codex available to all employees, expanding access beyond engineering teams.
Codex will help engineers explore codebases, propose changes and run tests. ChatGPT Work is intended for a broader range of tasks across the organization.
OpenAI, in turn, plans to adopt more ServiceNow workflows in its own operations. Both companies are positioning themselves as users of the integrated technologies they are developing.
Internal adoption may help identify practical deployment problems, but it does not establish a particular productivity improvement or financial return.
The companies have not released detailed pricing for all future integrations or independently verified cost savings associated with the expanded partnership.
Governance is the central challenge for enterprise agents
Enterprise AI is moving from assistance toward operational execution. Instead of merely recommending a next step, an agent can potentially carry out actions in systems that determine how an organization functions.
This raises the importance of permission boundaries, auditability and human oversight. Updating an incident record or triggering a business workflow can have consequences beyond the accuracy of a conversational answer.
ServiceNow's existing governance model provides an important foundation, but organizations will still need to examine how AI agents behave when instructions are ambiguous or external information is unreliable.
The more autonomous a workflow becomes, the more valuable it is to distinguish between information retrieved by an agent, decisions suggested by a model and actions actually executed by an enterprise system.
Performance should also be measured at the task level. The number of successful operations, required human interventions, access-control violations and completeness of audit records will matter more than simply counting model responses.
For now, the confirmed available component is the ServiceNow MCP plugin in the OpenAI marketplace. Enhanced capabilities are expected by the end of 2026, followed by the announced voice, security and Codex integrations during 2027.