Oracle Embeds OpenAI Into Its Business Software in What It Calls an Enterprise ‘Coming-Out Party’
Credit: Ken Yeung

Oracle has become one of OpenAI’s most important infrastructure partners. The AI company trained its GPT-6 Astra model at Oracle’s Abilene, Texas, data center site, the flagship campus of Stargate, the $500 billion data center venture OpenAI formed with Oracle, SoftBank, and MGX. Karan Batta, who runs product for Oracle Cloud Infrastructure (OCI), described the relationship as “a giant, very deep interoperable partnership.” But until now, the collaboration has centered on the computing power behind OpenAI’s models.

That relationship is expanding at this year’s Oracle AI World conference, where the two companies revealed they’re now building products together. On Tuesday, Oracle introduced four co-engineered tools that bring OpenAI’s models, ChatGPT, and Codex into Oracle’s applications, databases, and cloud services. Greg Brockman, OpenAI’s president and co-founder, called it an “expanded partnership” that will put frontier AI “into everyday operations.”

For OpenAI, the move creates another chance to bolster its credentials with business customers, potentially giving it a temporary edge over Anthropic. Batta views the new products as OpenAI’s formal introduction to the enterprise, despite OpenAI already releasing business tools. “This is truly [their] coming-out party for enterprise,” he declared. “OpenAI…has been very much the consumer-focused company, and that’s why these…products really matter.”

The four tools announced on Tuesday each target a different business need, from giving employees a single place to work across Oracle’s systems to finding security flaws before attackers do, building custom, secure apps, and helping small businesses get started with AI.

Oracle AI Workspace

The first product addresses a shift in how employees interact with software, which increasingly happens through chat. Salesforce, Adobe, ServiceNow, and other enterprise tech providers have embraced what the industry calls going “headless,” opening their platforms so AI agents can use them without relying on the companies’ own interfaces. It’s such a trend that Marc Benioff, Salesforce’s chief executive, recently called it the “year of interfaces.” Batta prefers a different label, saying that it’s more “agent-ready.”

Where Salesforce has transformed Slack into the new interface for AI and work, Oracle’s version runs through ChatGPT. Through a set of plugins, Oracle AI Workspace is designed to let employees describe what they need in plain language, and the assistant gathers information from across Oracle’s systems, completes approved workflows, and acts on their behalf. And it’s not just one app in Oracle’s portfolio—it’s designed to connect with Oracle Fusion Applications, Oracle NetSuite, Oracle AI Database, OCI, and other enterprise tools and data sources.

The approach raises an obvious question: Oracle doesn’t own a workplace hub the way Salesforce owns Slack, or Microsoft owns Teams. So why not build one instead of handing the front door to OpenAI? Batta contends the interface was never the hard part, since enterprise customers interact with Oracle’s systems largely through natural language as they look for clear answers to their problems. However, “those answers require you to make multiple API calls across different systems that interact with each other,” he said. Building a Slack-like app, in his view, wouldn’t make those answers any better.

This is where Batta reframes one of the industry’s favorite talking points. Software makers have frequently pitched their platforms as a cure for context switching, the constant jumping around between apps. Batta sees a bigger headache behind the screen: a supply chain application, database, and older Oracle software such as PeopleSoft or E-Business Suite (EBS) struggle to share information. “Having context switching across those systems is really, really hard,” he said, adding that customers want that problem solved and they don’t care what chatbot they use—ChatGPT or Claude. Batta summed up their attitude this way: “As long as you solve my big issue, which is not the problem of [picking] this tool versus that tool…[it’s] all these back-end systems having to talk to each other and passing information across each other.”

Batta offered a security scenario as an example. A manager could ask ChatGPT to list every storage bucket with open ports in the company’s Oracle cloud account and then match that list against the employees who have access to those buckets. Answering requires Workspace to query both the HR system and OCI and merge the results. Next to that work, Batta views the chat window as “the easy part.”

Oracle AI Security Mesh

Oracle and OpenAI have also built Oracle AI Security Mesh, a managed service designed to use OpenAI’s cyber models to find and fix vulnerabilities in a company’s applications and infrastructure, whether they run on OCI, another cloud, or on-premises. Oracle pairs those models with its own testing tools and infrastructure, independent validation, and operational safeguards it manages to reduce the risk of outages during testing. The result lets customers put advanced AI security testing to work without building and running their own AI systems to probe and defend their networks.

Batta pointed out that AI Security Mesh grew out of a tool Oracle first built for itself. “Oracle’s got software that we’ve been launching for 50 years, so we’re actually very, very good at taking the cyber models from OpenAI and finding the vulnerabilities, bugs, and then going and fixing them,” he said. To do that, Oracle built what Batta calls “a whole harness” around OpenAI’s models. The effort helped the company move from yearly software patches to quarterly and monthly. “Our software, just overall, just became more secure.”

Most companies lack the staff and expertise to do the same. “The long tail of enterprises don’t have that capability,” he added. “What we’re doing is we’re taking that harness along with the cyber models, and we’re putting that into a managed service.”

Oracle App Composer

Vibe coding has made it easy to spin up an app from a simple description, but Batta argues that the results often fall short of what a business needs. He pointed to OpenAI’s own Sites tool, which turns a prompt into a hosted website but leaves out identity, security, availability, and performance. Oracle App Composer is meant to fill that gap. Oracle pitches it as a way for employees to turn ideas into enterprise applications without coding expertise. Users build with Codex, and Oracle handles deployment and operations through OCI.

Oracle App Composer is designed to connect those applications to approved Oracle data and workflows and to OCI services such as Kubernetes Engine and Autonomous AI Database. “You essentially describe your application, website, whatever, and it gives you all of the other things that make it…much more resilient and available,” Batta explained. “And you can deploy it with all the bits and bobs that you would need.” When I compared it to Weebly on steroids for the enterprise, he agreed. “That’s exactly what it is,” he replied.

To make sure those apps meet enterprise standards, Oracle built APEXlang, a new language for describing applications. When employees describe what they want, the AI generates a short APEXlang definition of the app, and Oracle’s APEX engine runs it, incorporating security, scalability, and availability. Because the AI never writes the code that actually runs the app, companies don’t have to trust and secure large volumes of AI-generated code.

Oracle AI SMB Bundle

The expanded Oracle-OpenAI partnership culminates in the debut of the Oracle AI SMB Bundle. It’s a free-to-start package designed to level the playing field for small- to medium-sized businesses, giving them access to AI tools they can use to ideate products, run operations, and grow. These companies receive $12,000 in annual Oracle Cloud credits and access to ChatGPT Business, Oracle AI Workspace, Codex, Oracle App Composer, and OCI. NetSuite QuickStart is also included to help businesses with managing core operations such as invoicing and cash flow. 

“These systems are very much meant for enterprises,” Batta said. “But if you’re a mom-and-pop shop down the road running a pizza company, and you’ve got five employees, you still need all these systems.  But you don’t need the heavyweight systems that you get from a company like Oracle.”

Bundles like these aren’t new. Other hyperscalers have offered similar programs, including AWS Activate, Google for Startups, and Microsoft for Startups. But perhaps what sets Oracle’s offering apart is that it isn’t just for software startups—it also includes ordinary small businesses. That said, while Oracle’s AI SMB Bundle is free to start, the company hasn’t shared future costs or said how long the program will run. And there are “terms, conditions, and restrictions” that apply.

Not Exactly a Debut 

These four products show Oracle customers how AI can deliver real business outcomes while giving OpenAI a direct line to businesses that already run on Oracle software. Batta’s description of the announcement as OpenAI’s “coming-out party” for the enterprise is generous, given how much business OpenAI already does with companies. Its enterprise business already accounts for more than 40 percent of OpenAI’s revenue and could match its consumer business by the end of this year. In 2024, the company reported that 92 percent of Fortune 500 companies were using its products.

The partnership could also strengthen OpenAI’s pitch to investors as it pursues a public listing. Rival Anthropic set a benchmark for that comparison in its S-1 filing on Monday, disclosing that as of Sept. 30, 2026, it had more than 11,900 customers spending at least $100,000 a year, more than 1,900 spending at least $1 million, and 250 spending at least $10 million. Big-spending enterprise accounts like those make up much of Oracle’s customer base. By building its tools into the software that, in Oracle chief executive Clay Magouyrk’s words, businesses “rely on to run their most critical operations,” OpenAI gains a path to the kind of large contracts investors are likely to weigh when they compare the two companies.

Right now, all four run only on OpenAI’s technology. “All OpenAI powered,” Batta confirmed. That may not last, however. Oracle already offers models from several providers elsewhere in its portfolio, and Batta called it a “safe assumption” that these products could eventually support others, such as Anthropic’s Claude. He added a caveat, though: Workspace and Security Mesh are “so deeply integrated with the model itself” that switching wouldn’t be a simple choice.

Whether or not this counts as OpenAI’s coming-out party, Oracle is the one throwing it. By opening its applications, databases, and cloud to OpenAI, Oracle is betting that its software becomes more valuable when customers can reach it through the AI tools they already use. Batta framed it as a matter of choice: customers can still work with Oracle directly, “but you can also access it through systems like OpenAI.” The harder test comes next: Oracle hasn’t said when any of the four products will be available or what most of them will cost, and its own disclaimer warns that timing could change.

Disclosure: I attended Oracle AI World as a guest of the company, with travel and accommodation expenses covered. What I write reflects my own reporting and analysis, and no one reviewed or approved this post before publication. 

Discover more from Ken Yeung

Subscribe to get the latest posts sent to your email.