When OpenAI launched Frontier in February, the announcement positioned it as a platform for enterprise AI agents. What this really represented was a direct challenge to the revenue structure that had supported the software industry for the better part of two decades.
Frontier is designed to act as a semantic layer across an organization’s existing systems, connecting data warehouses, CRM platforms, ticketing tools, and internal applications, allowing AI agents to operate in the same business context as human employees. OpenAI calls these agents “AI coworkers” who can be onboarded, assigned identities, granted privileges, and reviewed for performance.
Early customers include Uber, State Farm, Intuit, and Thermo Fisher Scientific. The commercial ambitions behind the platform are not subtle. OpenAI CFO Sarah Friar said enterprise customers currently account for about 40% of the company’s revenue, and the company aims to increase this number to nearly 50% by the end of the year. The vehicle is Frontier.
What Frontier is actually doing for enterprise workflows
Frontier’s lawsuit is based on issues that the CIO has consistently described from 2025 to this year. The idea is that because agents are introduced separately, complexity is increased rather than removed. Each new agent becomes an integration point and requires its own data connectivity and governance controls, resulting in massive fragmentation.
OpenAI’s answer is shared business context. Rather than each agent building their own understanding of how your organization works, Frontier provides a centralized layer that all agents can see. Fidji Simo, CEO of Applications at OpenAI, made that clear in his launch briefing, citing his time running Instacart.
“We spent months integrating each of the tools we selected. We didn’t even really get what we wanted. Each tool was good for one use case, but they weren’t integrated and didn’t work with each other, so we just created more silos.”
The results OpenAI has cited from its early deployments are noteworthy. A global investment firm uses Frontier agents throughout its sales process, freeing up more than 90% of sales time previously spent on administrative tasks. One technology customer reported saving 1,500 hours a month on product development. For a major manufacturer, the agent reduced the production optimization process from six weeks to one day.
The frontier is also intentionally open. Manage agents built by OpenAI, agents built in-house by enterprise teams, and agents from third-party providers such as Google, Microsoft, and Anthropic. This openness is both a design principle and a positioning move. This expands the area that Frontier can control, while also making it difficult to ignore Frontier as a vendor lock-in strategy.
The seat license issue that no one wants to talk about out loud
The serious concerns of incumbents are structural. The per-seat licensing model that has made SaaS so profitable assumes that software usage scales with the number of employees. If an AI agent handles workflows that previously required a human employee to log into Salesforce, that seat license becomes less justifiable. Fortune explained it firsthand. The market’s concern is that platforms like Frontier will make SaaS software “invisible” and therefore less valuable.
Salesforce stock has fallen more than 27% so far this year, with analysts saying the company’s agent AI disruption is more of a concern than fundamental financial weakness. The company’s performance in the fourth quarter of FY2026 was strong. With revenue reaching $11.2 billion in the quarter, Agentforce’s annual recurring revenue reached $800 million, and the company closed 29,000 Agentforce deals.
Shares fell in after-hours after guidance was lower than Wall Street expected.
Incumbent companies are not standing still. Salesforce introduced something called the Agentic Enterprise License Contract, a fixed-price, all-you-can-use model for Agentforce that aims to make consumption more predictable for enterprise buyers.
ServiceNow moved to consumption-based pricing for some of its AI agent products and in January signed a multi-year agreement with OpenAI to build frontier model capabilities directly into its platform. Microsoft introduced pay-as-you-go pricing alongside Copilot Studio’s per-user model.
The axis of pricing is important. This shows that these companies understand that a seat licensing model cannot leave agent AI alone. The question is: will re-pricing be enough, or will the architecture itself need to change?
Two bets about where the intelligence layer should be located
The strategic divide in enterprise AI currently runs along a single fault line. In other words, should the AI agent reside within the system of record or on top of it? Salesforce and ServiceNow are betting on the built-in model. They argue that agents are most effective when they sit closest to the data, and that CIOs will more easily trust governance and compliance management from vendors who already manage their workflows.
Salesforce CEO Marc Benioff describes Agentforce as “the operating system for agent enterprises.” ServiceNow positions its AI Control Tower as a centralized governance layer for all agents, regardless of their origin.
Like OpenAI, Anthropic with Claude Cowork is also betting on overlay models. Rather than replacing existing systems, Frontier uses open standards to connect existing systems. The pitch is that enterprises don’t need to re-platform to run production-grade agents across their production environments.
There is merit to both arguments, and companies evaluating these platforms will find real trade-offs. An embedded approach provides tighter data control and faster time to value within a known ecosystem. The overlay approach provides flexibility and avoids the problem of agents only seeing data from one vendor.
What OpenAI doesn’t have is decades of institutional trust and existing contracts. OpenAI has the benefit of model capabilities, making it an increasingly credible argument that it can run the intelligence layer across an entire enterprise rather than just one product family.
What CIOs actually decide
Frontier is currently available to a limited number of customers, but will become more widely available in the coming months. Pricing has not been disclosed, and OpenAI is directing interested organizations to its enterprise sales team.
For CIOs, practical decisions are still not binary. Most large enterprises run Salesforce, ServiceNow, and Microsoft infrastructure simultaneously. The question at hand is whether Frontier will become an orchestration layer that connects these systems, or a competitive platform to replace them.
Denise Dresser, chief revenue officer at OpenAI, provided perhaps the most honest summary of the current position of enterprise AI agents. “What is really missing for most companies is an easy way to unleash the power of agents as teammates who can operate within the business without having to reinvent everything at its core.”
This gap is exactly what all platforms in this space are trying to fill. The difference with Frontier is that the companies making this claim now have the enterprise relationships, operational deployments, and model capabilities to back it up. SaaS incumbents have a head start when it comes to trust and data. Whether that proves sufficient is a central question for enterprise software for the rest of 2026.
(Photo courtesy of Austin Distel)
See also: Pushing OpenAI to the enterprise: The hidden story behind the AI sales race
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expos in Amsterdam, California, and London. This comprehensive event is part of TechEx and co-located with other major technology events such as Cyber Security & Cloud Expo. Click here for more information.
AI News is brought to you by TechForge Media. Learn about other upcoming enterprise technology events and webinars.
