For years, the promise of workplace AI has stopped at the same wall: the chatbot can explain your problem beautifully, but it cannot fix it. It cannot open the right ticket, check who may approve a change, or update the record where the work lives. On October 28, 2026, OpenAI and ServiceNow announced an expanded partnership to knock that wall down — connecting ChatGPT and Codex directly to the ServiceNow workflows that run IT, HR, security, and customer service inside large companies.
The headline feature is a new ServiceNow plugin for ChatGPT, available in the OpenAI marketplace now. With it, an employee can find similar past incidents, assess the risk of a planned change, and update a ServiceNow record without leaving the conversation. ServiceNow stays the system where the action is executed and governed; ChatGPT becomes the front door.
What the two companies actually announced
The October 28 announcement expands a multi-year agreement the companies struck on January 20, 2026, under which OpenAI models became a preferred intelligence capability on the ServiceNow AI Platform. The January deal put models like GPT-5.2 inside ServiceNow workflows, where they could summarize incidents and draft knowledge articles. The October expansion runs the connection in the other direction as well: ServiceNow workflows now show up inside ChatGPT and Codex.
According to ServiceNow’s announcement, the expansion has five moving parts:
- A ServiceNow plugin for ChatGPT, live in the OpenAI marketplace now, with richer capabilities expected by the end of 2026.
- OpenAI agents reaching ServiceNow workflows through Action Fabric, the layer that applies each customer’s business rules and permissions.
- Security and Risk solutions powered by OpenAI cyber models, entering early access in January 2027, with general availability planned for the second quarter of 2027.
- A Codex integration in ServiceNow’s AI Workflow Factory, expected in 2027, to help developers build applications on ServiceNow context.
- Voice-driven governance and risk assessment for ServiceNow’s Super Intelligence Control Tower, expected in the first quarter of 2027.
OpenAI is also joining ServiceNow’s newly announced Control Tower Alliance as a founding member, alongside companies including Anthropic, Amazon Web Services, Cisco, NVIDIA, and Okta. And both companies are deploying the tools on themselves: ServiceNow plans to give all of its employees access to ChatGPT Work and Codex, while OpenAI plans to run ServiceNow across its own operations.
Why “governed action” is the whole story
Most AI announcements celebrate reasoning — a smarter model, a longer memory, a better answer. This one celebrates permission. That distinction matters more than it sounds.
Inside a company, the hard part of a task is rarely knowing what to do. It is knowing who is allowed to do it, what approvals are required, and which system holds the authoritative record. ServiceNow already encodes those answers in role-based access, approval chains, and a system of record. The plugin’s job is to let ChatGPT act within those guardrails rather than around them.
As OpenAI’s own announcement of the partnership puts it, models paired with the ServiceNow workflow engine can access enterprise data, respect governance and permissions, and move work through approvals and updates until it is done.
The practical test is simple: if the AI suggests closing a ticket, does it check whether that employee may close it? ServiceNow’s design says yes — the available actions mirror each organization’s own configuration and access controls. That is also the limit worth remembering: the plugin is only as trustworthy as the permissions the customer has already set up. A company with sloppy access controls will not get careful AI by bolting on a plugin.
What is live now, and what is still coming
One of the easiest mistakes with partnership announcements is treating everything as shipped. Here, the timeline is unusually clear, and it is worth keeping straight:
| Capability | Status | What it means for you |
|---|---|---|
| ServiceNow plugin for ChatGPT (MCP plugin) | Available now in the OpenAI marketplace | Customers can connect ChatGPT to their own ServiceNow instance and choose which capabilities to expose. |
| Enhanced plugin capabilities | Expected by end of 2026 | Richer, interactive experiences inside ChatGPT, such as an on-call engineer reviewing alerts, changes, and affected services in one view. |
| Voice-driven risk assessment in the Control Tower | Expected Q1 2027 | A risk professional describes an AI use case aloud; the agent pre-fills the assessment, asks follow-ups, and assigns a risk rating. |
| OpenAI-powered Security and Risk solutions | Early access January 2027; general availability Q2 2027 | Joint tools to identify, triage, validate, and mitigate attack paths across code, apps, networks, and identities. |
| Codex in AI Workflow Factory | Expected in 2027 | Developers build ServiceNow applications with Codex, inheriting ServiceNow context and governance. |
The gap between “available now” and “coming in 2027” is the honest shape of this deal: a real, usable first step plus a roadmap. Businesses should plan around the plugin that exists today, not the security suite or Codex integration that are still dated for next year. Further coverage of the October 28 expansion confirms the same sequence: the plugin now, the broader integrations through 2027.
What it means for each team
IT service teams: fewer tabs, faster triage
The clearest day-one use case is incident work. Today, an IT agent typically searches past incidents in one tab, checks the change calendar in another, and updates the record in a third. The plugin collapses that into a conversation: find similar incidents, assess change risk, update the record. If that works as described, the time saved is not in typing; it is in context-switching, which is where incident response usually bleeds minutes.
Security teams: from 100,000 alerts to the ones that matter
ServiceNow’s announcement is candid about the security problem: a codebase scan can surface 10,000 to 100,000 vulnerabilities or more, and the hardest part is knowing which ones represent imminent risk. The planned joint solution aims to combine OpenAI’s cybersecurity capabilities with ServiceNow’s asset, access, and enterprise context to triage, validate, and mitigate attack paths. This is the piece to watch most skeptically — it is also the furthest from shipping, with general availability planned for mid-2027 — but the framing is right: security teams do not need more findings; they need defensible prioritization.
Developers: building apps where the governance already lives
The Codex integration in AI Workflow Factory targets a different bottleneck: rebuilding legacy applications as modern workflows. Because the apps inherit ServiceNow context and governance, developers get the business rules and permissions for free instead of re-implementing them. But it is a 2027 item, and coverage of the announcement notes the same timeline — so treat it as direction, not as a tool you can adopt this quarter.
HR and employee service: the quiet beneficiary
The January partnership already pointed at HR: employees asking “I need to view my benefits” in plain language, and getting action rather than just an answer. The October expansion extends that logic wherever ServiceNow runs employee workflows. For smaller companies, the lesson holds even without ServiceNow: the AI tools that save real time are the ones connected to the system where your business records live, not the ones with the cleverest chat window.
The bigger pattern: AI is moving into the system of record
ServiceNow is not the only vendor making this move, and that is the point. In the same month, ServiceNow also expanded its partnership with Google Cloud, connecting ServiceNow workflows into Gemini Enterprise. The pattern is unmistakable: model providers want their AI inside the software where business actually runs, and enterprise platforms want frontier models inside their governance. The winners will not be decided by benchmark scores but by whose permissions, audit trails, and approval chains hold up when an AI takes a real action.
There is a useful contrast with the other big enterprise AI story of recent weeks: companies blocking AI agents from their websites over security and control concerns. ServiceNow’s answer to the same anxiety is the opposite of blocking — structured access, where the agent can only do what the employee could already do, and every step is recorded. Whether customers trust that answer will depend on early deployments.
The Control Tower Alliance fits the same pattern: ServiceNow’s control plane for discovering, securing, governing, and measuring enterprise AI, with vendors — OpenAI among them as a founding member — agreeing to be governed, not just deployed.
A practical checklist before you switch it on
If your company runs ServiceNow, here is a sensible order of operations:
- Audit permissions first. Review who can close incidents, approve changes, and modify records. The plugin inherits whatever you have.
- Start with read-heavy use cases. Let teams find similar incidents and summarize records before enabling actions that change data.
- Pick one workflow to pilot. Incident triage is the natural candidate: measurable, high-volume, and easy to compare against the old process.
- Require human approval for irreversible actions. Keep approvals in the loop for production changes, access grants, and anything touching customer data.
- Log and review AI actions separately. Track which records the plugin touched, and review a sample weekly during the pilot.
- Train staff on what the plugin cannot see. It only knows what your ServiceNow instance exposes — not processes that live in spreadsheets and inboxes.
For businesses too small for ServiceNow, the checklist still applies to whatever tools you do use: connect AI to one governed system, pilot one workflow, and measure the result. Free, practical starting points are covered in our guide to the best free AI tools for students, many of which apply equally well to small business teams learning the ropes.
What to watch, and what to be skeptical about
Three honest caveats. First, no pricing, contract value, or implementation timeline was disclosed, so the cost of the plugin and the effort to deploy it remain open questions. Second, the most ambitious pieces — the security suite, voice risk assessment, and Codex integration — are all dated for 2027, and enterprise roadmaps slip. Third, the announcement leans on internal deployments at both companies as proof points; independent customer results will be the real test. Independent analysis of the deal draws the same line: the plugin is live, but the broader roadmap is still to come.
There is also a competitive question the announcement does not answer: if ServiceNow is building similar bridges to Google Cloud and OpenAI simultaneously, customers may eventually choose their model provider per workflow rather than per platform. That multi-model future is good for buyers, but it means the governance layer, not the model, becomes the durable asset. ServiceNow is betting its Control Tower is that layer.
Frequently asked questions
Is the ServiceNow plugin for ChatGPT available today?
Yes. According to the companies, the ServiceNow MCP plugin is available in the OpenAI marketplace as of October 28, 2026, with enhanced capabilities expected by the end of 2026. Customers connect ChatGPT to their own ServiceNow instance and choose which capabilities to expose.
Does ChatGPT get unrestricted access to my company’s ServiceNow data?
No — at least by design. ServiceNow says its Action Fabric applies existing business rules and permissions, so the actions available to an employee in ChatGPT mirror what that employee could already do inside ServiceNow. The practical safeguard is your own permission hygiene: audit roles and approvals before enabling write actions.
Do I need Codex to use this?
No. The ChatGPT plugin is the live piece. Codex integration in AI Workflow Factory is a separate, planned 2027 item aimed at developers building ServiceNow applications.
What does this mean if my business does not use ServiceNow?
Directly, very little today. Indirectly, it sets the template your vendors will follow: AI connected to a governed system of record, with permissions and audit trails intact. When evaluating any AI tool, ask the same questions ServiceNow’s customers should ask — what can it change, who approved that, and where is the record?
Is this the same as the January 2026 ServiceNow–OpenAI deal?
No. The January agreement made OpenAI models a preferred intelligence capability inside ServiceNow’s platform. The October 2026 expansion adds the reverse direction — ServiceNow workflows inside ChatGPT and Codex — plus the security, Codex, voice, and alliance pieces. January was AI inside ServiceNow; October is ServiceNow inside ChatGPT.
The bottom line
The OpenAI–ServiceNow expansion matters less as a product launch and more as a signpost. Enterprise AI is moving past the phase where a good answer was enough, into a phase where the AI is expected to do governed work — open the ticket, check the permission, file the record, and leave an audit trail. The plugin that shipped on October 28 is a modest first step, but it points in exactly the direction serious buyers have been asking for: less chat, more completed work, with the controls intact.
For your business, the action is not to rush. Audit your permissions, pilot one workflow, keep humans approving anything irreversible, and judge the rollout by completed work rather than demo quality. The companies that benefit first will be the ones that connect AI to one well-governed process, prove it works, and then expand from there.
Sources and Further Reading
- ServiceNow, “ServiceNow and OpenAI accelerate enterprise AI, from models and operations to cybersecurity and agentic development”, 28 October 2026. Official announcement of the expanded partnership. Title and date verified from the release itself on 19 November 2026.
- OpenAI, “OpenAI and ServiceNow expand partnership across enterprise workflows”, 28 October 2026. Official announcement of the expanded partnership. Title and date confirmed from secondary reporting of the OpenAI post on 19 November 2026.
- OpenAI, “ServiceNow powers actionable enterprise AI with OpenAI”, January 2026. Announcement of the multi-year agreement under which OpenAI models became a preferred intelligence capability on the ServiceNow AI Platform. The 20 January 2026 date is confirmed from secondary reporting; the OpenAI page itself, as fetched on 19 November 2026, shows no date.
- TokenPost, “OpenAI and ServiceNow Expand Partnership Across Enterprise Workflows”, 29 October 2026. Secondary reporting of the 28 October 2026 expansion.
- Unite.AI, “OpenAI and ServiceNow Expand Partnership With ChatGPT Workflow Plugin”, reporting the 28 October 2026 expansion. Secondary reporting; no publication date was shown in the page as fetched on 19 November 2026.
