Enterprise software has built its business model on the assumption that software consumption scales with human headcount. OpenAI‘s Frontier platform, launched in February, challenges that assumption directly.
According to the announcement, Frontier functions as a semantic layer connecting data warehouses, CRM platforms, ticketing tools, and internal applications, giving AI agents the same organisational context a human employee would carry. These agents are described as “AI coworkers” that can be on-boarded, assigned identities, granted permissions, and reviewed for performance. Early customers include Uber, State Farm, Intuit, and Thermo Fisher Scientific.
The architecture addresses a specific failure mode. When agents are deployed in isolation, each one requires its own data connections and governance controls, producing fragmentation rather than efficiency. Frontier’s answer is a centralised business context that every agent references, regardless of whether it was built by OpenAI, an internal enterprise team, or a third-party provider. That openness to external agents is both a design principle and a competitive tactic — it makes the vendor lock-in objection harder to sustain and broadens the surface area the platform can govern.
Fidji Simo, OpenAI‘s CEO of Applications, drew on her experience at Instacart to explain the problem the platform targets. “We spent months integrating each of the ones that we selected,” she said at the launch briefing. “We didn’t even get what we actually wanted, because each tool was good for one use case, but they weren’t integrated or talking to one another, so we were just reinforcing silos on silos.”
The performance figures cited from early deployments are specific. A global investment firm reported that Frontier agents in its sales process freed more than 90% of salesperson time previously consumed by administrative tasks. A technology customer saved 1,500 hours a month in product development. At a major manufacturer, agents compressed a production optimisation process from six weeks to a single day.
The Seat-Licence Problem
The threat to incumbent software vendors is structural, not competitive in the conventional sense. Per-seat licensing is profitable precisely because software use has historically mapped to headcount. If an AI agent handles the workflow that previously required an employee logging into Salesforce, the justification for that seat licence erodes. The report notes that Fortune described market fear of platforms like Frontier making SaaS software “invisible” and therefore less valuable. Salesforce‘s stock has declined more than 27% this year, a move analysts have attributed more to agentic AI disruption fears than to any weakness in fundamentals — the company reported $11.2 billion in quarterly revenue, $800 million in Agentforce annual recurring revenue, and 29,000 Agentforce deals closed, with the stock declining after guidance fell short of Wall Street’s expectations.
The incumbents are repricing. Salesforce has introduced what it calls the Agentic Enterprise License Agreement, a fixed-price model for Agentforce designed to make consumption predictable. ServiceNow has moved to consumption-based pricing for some AI agent offerings and signed a multiyear agreement with OpenAI in January to embed frontier model capabilities directly into its platform. Microsoft has introduced consumption-based pricing alongside its per-user model for Copilot Studio.
Repricing, however, does not resolve the deeper architectural question of whether AI agents belong inside systems of record or above them. Salesforce and ServiceNow argue the embedded model wins on governance and proximity to data. Marc Benioff has called Agentforce the “operating system for the agentic enterprise.” OpenAI CFO Sarah Friar has stated that enterprise customers currently represent roughly 40% of the company’s revenue, with a target of closer to 50% by year-end, and has identified Frontier as the primary vehicle for closing that gap.
Photo by Vitaly Gariev on Unsplash
This article is a curated summary based on third-party sources. Source: Read the original article
