With its new Frontier platform, announced earlier this month, OpenAI is looking to position itself as the best option for managing enterprise AI agents. But to succeed, the company that delivered the popular ChatGPT in 2022 will need to show it can manage the complexity of large-scale agent deployments — and do so better than  other tech vendors with the same aims.

Frontier is an “end-to-end” platform designed to help “enterprises build, deploy and manage AI agents,” according to OpenAI. It connects agents to core business systems — such as CRM, ERP, and data warehouses — and centralizes how these agents are configured, monitored and governed. 

Early customers include State Farm and Thermo Fisher Scientific.

Eric Goodness, Gartner distinguished vice president analyst, described it as an “intelligence layer” or “operating system” for enterprise AI agents. “It’s a full-stack approach by OpenAI to abstract away a lot of the complexity of middleware and other challenges, to create broad generative and agentic AI-types of solutions,” he said. 

“Enterprises are struggling to get from prototype and [proof of concept] and single step agents to a truly multi-agent, end-to-end flow,” said Tim Law, research director at IDC. The launch of Frontier “bodes well for enterprises who are looking to work with OpenAI,” he said.

OpenAI already has a presence in large organizations with an enterprise-friendly version of ChatGPT, as well as APIs and developer tools. It’s been “pivoting towards the enterprise over the past couple years,” Goodness said, partnering with software vendors and global systems integrators to sell generative AI (genAI) tools and services to the enterprise. 

But expanding from model provider to a control plane for agents brings with it another set of expectations. As Frontier moves closer to core business systems, enterprises are likely to scrutinize OpenAI’s long-term enterprise focus more closely, as well as its technical capabilities. 

Despite OpenAI’s early traction, there have been “some doubts about how committed they are to the enterprise stack,” said Law, with expansions into new areas raising questions questions about its focus. 

“There has been some sentiment that OpenAI was maybe distracted by other consumer-focused business launches,” he said, referring to reports that the company is developing AI hardware in conjunction with former Apple designer Jony Ive. He also pointed to involvement in large-scale data center buildouts to meet growing demand for AI. 

While OpenAI has cutting-edge technology and significant organizational capacity, it remains less established in enterprise environments compared to long-standing incumbents. “A lot of these applications and data management systems — especially from the hyperscalers — have been in place in organizations for years. They’re a known entity, a known quantity,” he said. “OpenAI has to ensure it can communicate and position itself not as just a cool new AI capability that’s growing exponentially in terms of its reasoning capabilities.”

With the launch of Frontier, OpenAI will be competing with enterprise software vendors that also want to position themselves as the orchestration layer for AI agents. Microsoft, for instance, last year announced its Agent 365 platform; it has some overlapping functionality with Frontier, focusing on agent governance and security, in particular, as well as multiple agent builder tools. Google, which has spent years pushing into the enterprise, also sells various products to build and manage agents. Both benefit from tight integration with their application suites, said Law.

Few large organizations have yet adopted agents at scale, but well-established  vendors could start with an advantage if businesses begin to invest more. When enterprise organizations evaluate large language models (LLMs) and agentic platforms, their “first preferred choice of provider is, in fact, the hyperscalers,” said Law.

While OpenAI has benefitted from its first-mover advantage in the enterprise, the landscape is changing quickly and “in 2025 a lot of the competitors caught up pretty quickly,” said Law. “I would count Anthropic among those….” That company “made a very well-defined and structured ‘go to market’ around enterprises. They won some significant market share in [the] enterprise based on their constitutional AI and their work focus around alignment and safety.”

Another challenge for OpenAI will be demonstrating a long-term commitment to specific industries. Though it has signaled its intentions with initiatives such as OpenAI for Healthcare — reflecting a concerted push into regulated sectors — gaining enterprise confidence will take time. 

“One of the outstanding questions to OpenAI and some others is, ‘How focused are you on my business?’ Because that is a primary driver for long-term relationships in vertical industries,” said Law.

Enterprises want the dependability of a long-term partnership, he added, including  assurances that new technologies can “integrate harmoniously into a heterogeneous landscape” that includes legacy systems, and “play nice in the sandbox with all the other providers.

“It’s not an easy thing to create a whole vertical strategy overnight,” Law said. Vendors such as SAP, Oracle and Salesforce have taken years to build deep industry capabilities. Developing that expertise requires domain specialists, intellectual property and long-standing systems integrator partnerships, “so that’s going to take them some time.”

It also will require continuing to build services and ecosystems needed to support agents at scale across complex environments. “OpenAI, at its core, is a model and intelligence company — they’re not a global systems integrator,” Goodness said. 

Scaling in the enterprise will require further partnerships with systems integrators and managed service providers. “To be able to scale moving forward — just as Anthropic is and all the other challengers — they’re going to have to have a rock solid, maybe certified, cohort of GSIs and managed service providers to help their customers create stable, unique types of solutions. And everybody’s starting at stage one right now in those terms,” said Goodness.

Along those lines, the company reportedly plans to expand its teams of forward deployed engineers — the specialized consultants that help roll out its services to large enterprise clients. 

That should help as it seeks to deploy products such as Frontier to top-tier clients, in particular, said Goodness. 

OpenAI has also announced a string of executive hires. “They’ve been bolstering the executive suite,” said Law. This includes a new CTO of applications who has “very much an enterprise focus, and you can see that starting to permeate the organization.” 

Law also pointed to OpenAI’s hiring of Anthropic’s AI safety and alignment lead, demonstrating to enterprises that OpenAI has an “equal focus” in relation to its rival. While the company has invested significantly in its safety and alignment work as part of its research, the hire signals renewed emphasis on safety and security readiness.

“That’s still an open question for lots of enterprises: ‘Can I be assured that I’m not significantly expanding my threat surface by implementing multi-agent, end-to-end workflows?’” said Law.

Other hires include former Slack CEO and Salesforce exec Denise Dresser, who joined as chief revenue officer; and Peter Steinberger, the creator of OpenClaw, who opted to join OpenAI rather than build his own business. 

Some of these moves have “reassured enterprises,” said Law. 

Of course, OpenAI is not alone in needing to convince customers it can provide the layer for managing and deploying agents at scale. The broader market is just taking shape, and no vendor can claim to have achieved widespread agent adoption in the enterprise.

“It’s the wild, wild west right now; everybody has challenges,” said Goodness.