OpenAI and ServiceNow Expand Partnership: ServiceNow Inside ChatGPT, Codex Inside ServiceNow

OpenAI and ServiceNow are widening their January partnership into a two-way integration. In an announcement dated October 28, 2026, OpenAI says ServiceNow workflows will be reachable from inside ChatGPT, OpenAI models and Codex will power more of ServiceNow's products, and every ServiceNow employee will get ChatGPT Work and Codex.

Only one piece is live: a ServiceNow MCP plugin in the OpenAI marketplace. Voice, a joint security product and Codex inside ServiceNow's AI Workflow Factory are planned for between the end of 2026 and 2027. This post separates what ships now from what is a plan, and explains what it means if your company runs ServiceNow.

TL;DR: what ships and when

What ServiceNow in ChatGPT does

A stage with floating feature cards, standing for the new OpenAI ServiceNow partnership features rolling out through 2027

The first direction is ChatGPT reaching into ServiceNow. OpenAI describes an integration that brings "business context and workflow actions directly into ChatGPT", with this example: an IT team member can find similar incidents, assess the risk of a change and update a ServiceNow record without leaving ChatGPT.

Control sits with the customer. Organizations connect ChatGPT to their own ServiceNow instance and pick which capabilities to expose. ServiceNow's Action Fabric then applies existing business rules and permissions, so what a person can do through ChatGPT mirrors what their ServiceNow role already allows. That design choice matters for agents: the model does not get a new, wider permission set, it inherits the one the platform already enforces.

ServiceNow is also previewing interactive experiences inside ChatGPT. In OpenAI's example, an on-call engineer reviews linked monitoring alerts, recent changes and affected services, then assigns the right team and works toward a fix, all within ChatGPT. This builds on the interactive answers we described in ChatGPT's intelligent UI explainer.

The delivery mechanism is the Model Context Protocol. If MCP is new to you, start with our guide to MCP. The practical point is that the plugin is a standard connector, available now, rather than a custom deal-only integration.

What OpenAI models do inside ServiceNow

The second direction is OpenAI inside ServiceNow, in three planned areas:

1. Application development. A planned Codex integration in AI Workflow Factory will help developers build applications using ServiceNow's business context and platform capabilities. OpenAI models will also power Build Agent and Autonomous Engineer for application development and modernization.
2. Cybersecurity. A joint solution will pair OpenAI's cyber capabilities with ServiceNow's asset, identity and business context to investigate vulnerabilities, prioritize risks and coordinate remediation.
3. Voice. In the Super Intelligence Control Tower, people will describe an AI use case aloud and the system will fill in assessment fields and ask follow-up questions for what is missing.

"Super Intelligence Control Tower" is ServiceNow's name for its governance layer. Bill McDermott, ServiceNow's chairman and CEO, is quoted in the announcement describing ServiceNow as the place that "orchestrates, secures, and governs AI across the enterprise". Greg Brockman, OpenAI's president, says the aim is to help teams "build software, address security risks and put AI to work across their organizations."

For how ServiceNow is already positioning governed agents, see our report on ServiceNow Arc going generally available with AI Passport and NVIDIA OpenShell.

ChatGPT Work and Codex for every ServiceNow employee

ServiceNow plans to give all employees ChatGPT Work and Codex, beyond its engineering organization. Engineers will use Codex to explore codebases, make changes and run tests; other teams will use ChatGPT Work day to day. ServiceNow also plans to use OpenAI capabilities in its internal security work, and OpenAI says it uses ServiceNow workflows in its own operations.

This follows a run of company-wide rollouts. Uber put ChatGPT Work in front of all its staff after a pilot (our Uber coverage), and OpenAI expanded its Oracle relationship a day earlier (Oracle and OpenAI: ChatGPT gets Oracle apps). If you are unsure how the two products divide the work, our ChatGPT Work vs Codex guide explains it.

Where this fits: the January deal

The announcement says it "builds on our partnership announced in January." Secondary coverage of that January deal describes a multi-year partnership in which ServiceNow uses OpenAI models, with plans for speech-to-speech voice and computer-use models (Constellation Research summary). We have not independently verified the January terms and the October post does not restate them.

What is new in October is the reverse direction (ServiceNow inside ChatGPT), the Codex and security plans, and employee-wide access. That pattern, an enterprise platform exposing its actions to a general assistant, is the same one Oracle announced the day before.

What this means if you run ServiceNow

• Try the plugin first. It is the only live piece. Test it on a non-production instance and review which capabilities you expose.
• Audit your permissions. Since Action Fabric reuses existing rules, any overly broad ServiceNow role becomes an overly broad ChatGPT capability. Tighten roles before connecting.
• Log agent actions. Updating records from a chat is convenient and also a new write path. Keep audit trails; our OpenShell safety platform coverage shows what runtime controls look like.
• Do not plan around 2027 features yet. Security early access in January 2027 and Codex in AI Workflow Factory in 2027 are stated intentions, not shipped products.

A rollout checklist for IT and security teams

A plugin that can write to your system of record deserves the same review as any new integration. A sensible order of work, based on the design OpenAI describes:

1. Inventory roles. List which ServiceNow roles would be able to reach ChatGPT. Since Action Fabric reuses permissions, the role list is effectively the capability list.
2. Start read-heavy. Expose search and summarization first, such as finding similar incidents, before enabling record updates. OpenAI says customers choose which capabilities are available, so you can phase this.
3. Separate test and production. Connect a development instance, run real incident scenarios, and compare what ChatGPT proposes with what an analyst would do.
4. Define a human checkpoint. For changes with risk, such as approving a change request, require a person to confirm even if the connector technically allows the action.
5. Review logging. Decide where the record of a chat-initiated action lives, in ServiceNow's audit trail, in ChatGPT's workspace history, or both.
6. Revisit when features land. The security product and Codex integration arrive in 2027 and will change the risk picture; treat each as a new review.

How this compares with other enterprise assistant deals

The shape of this deal is now familiar. A system-of-record vendor exposes actions to a general assistant, and the assistant vendor embeds its models back into the vendor's product. Oracle's announcement a day earlier followed the same two-way logic for ChatGPT and OCI, and ServiceNow's own agent, Arc, shows the competing instinct to keep the assistant inside the platform. The open question for buyers is which surface their staff will actually use: the vendor's console or the assistant they already open all day. By putting a connector in the OpenAI marketplace now, ServiceNow is hedging in favor of both.

What the announcement does not say

• No pricing, contract value or seat counts.
• No benchmark or measured time saving; the use cases are examples, not results.
• No detail on data handling for records that pass through ChatGPT, beyond the permission model.
• No list of which ServiceNow capabilities the plugin exposes.

Those gaps are normal for a partnership post, but they are what an IT or security team will ask first. Expect more detail when the enhanced capabilities arrive at the end of 2026.

Questions people will ask

Is this OpenAI replacing other models inside ServiceNow? The post does not say that. It describes OpenAI models powering specific products such as Build Agent and Autonomous Engineer; it makes no exclusivity claim.

Does ChatGPT get more access than a human user would? According to OpenAI, no: Action Fabric applies the organization's existing business rules and permissions.

Is voice available now? No. Voice in the Control Tower is expected in the first quarter of 2027.

Related reading

• Oracle and OpenAI: ChatGPT Gets Oracle Apps
• ServiceNow Arc Is Generally Available
• Uber Brings ChatGPT Work to All Employees
• ChatGPT Work vs Codex: Complete Guide
• What Is MCP? Model Context Protocol Guide
• OpenAI Dev Day 2026
• Source: OpenAI, OpenAI and ServiceNow expand partnership across enterprise workflows

Accurate as of October 28, 2026, based on OpenAI's announcement. Dates for planned features may change.