OpenAI Frontier: The Challenge That Could Disrupt the SaaS Industry

OpenAI Frontier: A Revolution for Enterprise AI Agents

Last February, OpenAI made waves by unveiling Frontier, a platform dedicated to AI agents for businesses. While presented as a mere technological innovation, this announcement has actually highlighted a major challenge for the traditional business model of the software industry.

Frontier positions itself as a semantic layer within existing enterprise systems. Its role is to connect various elements such as data warehouses, CRM platforms, ticketing tools, and internal applications. The goal is to enable AI agents to operate with a level of understanding and business context equivalent to that of a human employee. OpenAI refers to these agents as “AI collaborators,” capable of being integrated, endowed with identities, permissions, and performance evaluations.

Among the early adopters of this technology are giants like Uber, State Farm, Intuit, and Thermo Fisher Scientific. According to Sarah Friar, CFO of OpenAI, enterprise clients currently represent about 40% of the company's revenue. She aims to increase this proportion to nearly 50% by the end of the year, with Frontier as the main lever to achieve this goal.

Integrating Frontier into Business Processes

One of the major arguments in favor of Frontier is its ability to simplify processes by eliminating the complexity added by agents deployed in isolation. Each new agent represents an integration point requiring its own data connections and governance controls, which can lead to increased fragmentation. OpenAI offers a solution through a shared business context, allowing all agents to refer to a centralized layer rather than developing their understanding of the organization individually.

Fidji Simo, CEO of Applications at OpenAI, shared her experience at Instacart, where integrating various tools proved to be tedious. “We spent months integrating each of the tools we had selected. We didn’t even get what we really wanted, as each tool was good for one use case, but they weren’t integrated or communicating with each other, so we were just reinforcing silos on top of silos.”

The results from the initial deployments of Frontier are promising. A global investment firm used these agents for its sales process, freeing up more than 90% of the time that salespeople previously spent on administrative tasks. A tech client reported saving 1,500 hours per month in product development. In a large manufacturing company, agents reduced a production optimization process from six weeks to just one day.

Frontier allows for the management of agents built by OpenAI, by internal enterprise teams, as well as those from third-party vendors. This openness is a fundamental principle of its design and positioning, making Frontier difficult to dismiss based on vendor lock-in and expanding the surface it can govern.

Challenging the Per-User Licensing Model

The arrival of Frontier raises structural concerns for established SaaS players. The per-user licensing model, which has long ensured the profitability of SaaS, is based on the idea that software usage is proportional to headcount. However, if an AI agent can manage the workflow previously performed by a human employee, the justification for this user license weakens. Fortune has described the market's fear that models like Frontier will render SaaS software “invisible” and consequently less valuable.

Salesforce's stock has dropped by more than 27% this year, a decline that analysts attribute more to fears of disruption from agentic AI than to weaknesses in its financial fundamentals. The company's revenue reached $11.2 billion during the quarter, with annual recurring revenue from Agentforce hitting $800 million and 29,000 Agentforce contracts signed. Despite this, the stock fell after forecasts fell short of Wall Street expectations.

In the face of these challenges, established players are not remaining idle. Salesforce has introduced the Agentic Enterprise License Agreement, a fixed-price, all-inclusive model for Agentforce, aimed at making consumption more predictable for enterprise buyers. ServiceNow has adopted consumption-based pricing for some of its AI agent offerings and signed a multi-year agreement with OpenAI to integrate Frontier model capabilities directly into its platform. Microsoft has also introduced consumption-based pricing with its per-user model for Copilot Studio.

These pricing adjustments show that companies understand the per-user licensing model cannot survive agentic AI without change. The question is whether the price adjustments are sufficient or if the architecture itself needs to evolve.

Divergent Visions on the Position of AI Agents

A crucial question arises: should AI agents be integrated within reference systems or overlay them? Salesforce and ServiceNow are betting on the integrated model, asserting that agents are more effective when they are closest to the data, and that CIOs will trust governance and compliance controls coming from vendors already managing their workflows.

Marc Benioff, CEO of Salesforce, described Agentforce as the “operating system of agentic enterprise.” ServiceNow positions its AI Control Tower as a centralized governance layer for all agents, regardless of their origin.

OpenAI, and to a lesser extent Anthropic with Claude Cowork, are betting on the overlay model. Frontier sits above existing systems, using open standards to connect them. The argument is that companies should not have to change platforms to make production-quality agents work in their operations.

Both approaches have their merits, and companies evaluating these platforms will face real trade-offs. The integrated approach offers better data control and faster time to value in a known ecosystem. The overlay approach offers flexibility and avoids the issue of agents only being able to see data from a single vendor.

What established players have that OpenAI lacks is decades of institutional trust and existing contracts. What the AI leader possesses is the advantage of model capability and an argument that it can operate the intelligence layer across the enterprise.

Frontier is currently available to a limited number of clients, with broader availability expected in the coming months. Pricing has not been disclosed, with OpenAI directing interested organizations to its enterprise sales team.

Many large companies simultaneously use the infrastructure of Salesforce, ServiceNow, and Microsoft. The immediate question is whether Frontier will become an orchestration layer that connects systems or a platform that replaces them.

OpenAI's Chief Commercial Officer, Denise Dresser, stated, “What most companies still lack is simply a straightforward way to unleash the power of agents as teammates who can operate within the enterprise without needing to rework everything underneath.”

Each platform in this space claims to bridge the gap. Established SaaS players have an advantage in trust and data. The central question for enterprise software until the end of 2026 is whether this will prove sufficient.