OpenAI’s Enterprise AI Signals Point to Deeper Workflows, Not Just More Usage
OpenAI’s enterprise materials suggest that leading firms gain an advantage by embedding AI more deeply in day-to-day work, while Codex expands role-specific workflow tooling.
OpenAI’s latest enterprise materials point to a widening difference between organizations that use AI occasionally and those that build it into everyday work. Its public B2B Signals data indicates that frontier firms use 3.5 times more AI intelligence per worker than typical firms on average, up from 2 times a year earlier. Separately, OpenAI has expanded Codex with role-specific plugins, Sites and annotations, signaling a product direction centered on more delegated, tool-connected knowledge work.
That evidence supports a more useful conclusion than the specific adoption figures circulating in the original signal. OpenAI’s public documentation does not substantiate the exact claim that the top 10% of enterprises use plugins twice as often and skills six times as often as typical firms. What it does document is a substantial frontier-versus-typical gap in the depth and intensity of AI use, alongside a broader plugin-enabled workflow toolkit for Codex.
What OpenAI’s enterprise data shows
OpenAI’s State of Enterprise AI 2025 report and its May 2026 B2B Signals materials frame enterprise adoption as more than a question of how many employees have access to an AI assistant. The relevant distinction is whether employees use AI repeatedly, across meaningful work, and with enough organizational support to make that usage part of normal operations.
The B2B Signals overview describes frontier organizations as using 3.5 times more intelligence per worker than typical firms, compared with a 2-times gap a year earlier. The reported trend suggests that the separation is growing as advanced users move beyond basic prompting into sustained, role-specific work. OpenAI also identifies usage patterns that are markedly higher in particular contexts, including 2 times more messages per seat, 16 times more Codex messages, and 6 times more data-analytic messages. These are contextual measures, not a universal measure of every enterprise feature.
| Public OpenAI signal | Earlier or typical reference | Current documented indication |
|---|---|---|
| AI intelligence per worker at frontier firms | 2 times typical-firm usage a year earlier | 3.5 times typical-firm usage on average |
| Message activity | Typical-seat baseline | 2 times more messages per seat in the cited context |
| Codex and data-analysis activity | Relevant comparison baseline | 16 times more Codex messages and 6 times more data-analytic messages in cited contexts |
The table also shows why broad claims about plugin and skill frequency need care. OpenAI’s published figures use different measures and contexts, including percentiles, per-worker intensity and message activity. They demonstrate that frontier firms are operating differently, but they do not establish the precise plugin and skill multipliers stated in the original signal.
For business leaders, the practical lesson is not simply to increase chat volume. Higher usage can reflect genuine workflow integration, but volume alone does not show whether an organization has improved quality, speed, accountability or decision-making. The stronger question is where AI is connected to a defined job, a repeatable process and an accountable owner.
Codex expands the workflow layer
OpenAI’s June 2, 2026 announcement, Codex for every role, tool, and workflow, describes a set of additions intended to make Codex more usable across functions. The release includes six role-specific plugins that collectively bundle 62 apps and 110 skills, as well as Sites and annotations.
The role-specific approach matters because it shifts the framing from a general-purpose coding assistant toward a configurable work environment. Rather than asking each employee to assemble a workflow from scratch, organizations can potentially give teams a more structured starting point tailored to their responsibilities. That can reduce friction, but it also raises the importance of thoughtful implementation.
The documented changes include:
- Six role-specific Codex plugins.
- A bundle spanning 62 apps and 110 skills.
- Sites and annotations as part of the expanded Codex workflow experience.
- A focus on roles, tools and workflows rather than a single generic use case.
For enterprises, plug-in-enabled AI introduces governance questions alongside productivity opportunities. Access to connected tools should be mapped to job responsibilities. Leaders need to determine which information can be accessed, which actions require review, how outputs are checked, and who is responsible when AI-assisted work affects customers, code, analysis or internal decisions.
This is where the frontier-firm data and the Codex release intersect. Organizations that derive more value from AI are likely not treating it as an isolated interface. They appear to be making it available within recurring work, while the newer Codex tooling offers a way to organize those connections around roles. The public materials do not disclose each company’s implementation model or return on investment, so the evidence should not be read as proof that any plugin deployment automatically produces superior results.
A more disciplined adoption program should therefore prioritize a small number of high-frequency workflows, define success measures before rollout and establish controls appropriate to the data and actions involved. It should also distinguish between experimentation and production use. A successful pilot may show employee interest, while a durable deployment must show that it can operate reliably within the organization’s security, quality and accountability requirements.
Businesses moving from isolated AI trials to connected workflows need a clear operating model for governance, role design and measurement. Scalevise helps teams turn AI ambitions into practical implementation plans, identifying where automation and connected tools can create measurable value without losing control of critical processes. A focused assessment can clarify which workflows are ready, what safeguards they require and how results should be evaluated. Discuss an AI consultancy project with Scalevise to map your next enterprise AI initiative.
Frequently Asked Questions
What does OpenAI mean by frontier firms?
OpenAI uses the term for organizations at the leading end of enterprise AI adoption and usage. Its public B2B Signals materials emphasize their greater AI intensity per worker relative to typical firms.
Did OpenAI confirm that top enterprises use plugins twice as often and skills six times as often?
No. The supplied OpenAI materials do not publish those exact across-enterprise plugin and skill figures. They document other frontier-usage measures, including 3.5 times more intelligence per worker on average.
What did OpenAI add to Codex?
OpenAI described six role-specific plugins bundling 62 apps and 110 skills, plus Sites and annotations, in its June 2, 2026 Codex workflow announcement.
Why do role-specific plugins matter for enterprise AI adoption?
They can give teams a more structured way to connect AI with recurring work and relevant tools. Enterprises still need to define permissions, review processes and accountability for connected workflows.
Conclusion
OpenAI’s public data suggests the enterprise AI gap is increasingly about workflow depth, not access alone. Its Codex additions provide more role-oriented building blocks for that shift, while the published evidence also reinforces the need to separate documented adoption signals from unsupported feature-frequency claims. The organizations most likely to benefit will pair connected AI workflows with clear governance, measurable objectives and responsible operational ownership.