OpenAI just made its boldest enterprise move yet with the launch of Frontier, a new end-to-end platform for building, deploying, and managing AI agents.
Frontier functions as "AI coworkers" inside real business workflows, according to OpenAI. Early users include Uber, State Farm, Oracle, Cisco, Intuit, and Thermo Fisher.
This platform indicates OpenAI wants to be the central nervous system for how AI agents operate inside every enterprise. And that puts it on a direct collision course with Microsoft, its biggest investor.
To look at what this means for the future of enterprise software, I talked with SmarterX and Marketing AI Institute founder and CEO Paul Roetzer on Episode 196 of The Artificial Intelligence Show.
Frontier is built around four pillars:
Agents built on Frontier can follow users into ChatGPT, browse the web, and operate applications, including Excel and PowerPoint. The platform launched alongside GPT-5.3-Codex, which is 25% faster and set new benchmark records (SWE-Bench Pro).
Roetzer sees this as one of the most significant OpenAI launches.
"This is one of those where you probably look back in six months and think, ‘Oh, OK, they were telling us exactly what they were going to do,'" he says.
He lays out the core requirements for AI coworkers to actually work: They need to understand how work gets done across systems. They need access to tools to plan, act, and solve real problems. They need to understand what good looks like. And they need an identity, permissions, and boundaries that teams can trust.
Frontier is OpenAI's answer to all of these requirements on a single platform.
One of the most important things Frontier addresses for enterprises is the growing headache of agent sprawl.
As more teams deploy agents in isolation the complexity increases. You've got agents scattered everywhere with no centralized way to manage permissions, data access, or quality.
Frontier's approach is centralized context plus execution plus evaluation plus governance. Each new agent, whether it's from OpenAI, built in-house, or from a third party (Google, Microsoft, or Anthropic), reduces the friction of integrating into what Roetzer describes as "a team of agents."
Critically, it doesn't force companies to get rid of their existing software stack. Frontier works with the systems teams already have, integrating across existing applications.
This is where things get potentially disruptive for SaaS companies such as HubSpot, Salesforce, Workday, and ServiceNow in three crucial areas.
First, if work increasingly happens through agents inside ChatGPT, users spend less time in native software. You don't need to log into HubSpot if your CRM data is flowing into your AI workspace.
"Now imagine if I just create an agent and that agent goes and does everything in HubSpot with one seat," says Roetzer. "Maybe I have to buy a seat for my agent, but that agent can go do the work of 25 people. And now you realize why seat licenses collapse."
Secondly, the strongest position belongs to the app that owns the authoritative data, which is good news for HubSpot and Salesforce. But if OpenAI becomes the primary work surface, those platforms become databases under the hood, not places humans log into every day.
Finally, pricing pressure accelerates. As agents automate tasks, users will push back on per-seat pricing when humans aren't doing the work.
Roetzer believes the pricing question leads to an uncomfortable but inevitable conclusion: AI agents begin to replace human workers.
"As much as that sounds like a horrible thing, it's the most logical thing," he says. "I know the value I get out of employing a human for a year. So what is the value of employing a collection of agents for the year?"
The math is simple. If an agent can replace five people, the ROI goes up.
"Not the thing people want to hear, but trust me, that is how they're thinking," says Roetzer.
Frontier also creates an awkward dynamic between partners OpenAi and Microsoft. Frontier puts OpenAI in direct competition with Microsoft Copilot for enterprise software dollars.
Microsoft's Commercial CEO Judson Althoff circulated internal talking points to the company's sales organization after Frontier's announcement, according to reporting from The Information. The core message: Don’t panic. Microsoft is uniquely positioned.
At the same time, OpenAI is no longer just a model provider sitting underneath Microsoft's products. It's building its own enterprise platform and several large enterprises are already hedging their bets between the two.
Roetzer flagged one particular detail from OpenAI's Frontier launch post: a chart breaking down the different layers of the platform.
"There's a chart on the launch post that I think probably ends up being a pretty important chart about how they see the future," he says.
If Frontier delivers on even half of its promise, the implications are massive not just for how companies deploy AI, but for the entire economics of enterprise software. Seat-based pricing, standalone SaaS interfaces, even the concept of "logging in" to do your job, all could end.