Most companies are still treating AI like a smarter search bar. A small number are doing something fundamentally different — and according to OpenAI’s new B2B Signals research, the gap between those two groups is already wide enough to matter. The report, published on May 6, 2026, draws on real usage data and enterprise interviews to map out what separates surface-level AI adoption from the kind that actually builds competitive advantage. The short version: the firms pulling ahead aren’t using more AI tools, they’re changing how work gets done at a structural level.
What Is B2B Signals and Why Did OpenAI Publish It?
OpenAI doesn’t usually publish detailed research about how its enterprise customers use its products. So the decision to release a formal B2B Signals report is itself a signal worth paying attention to. This isn’t a whitepaper designed to justify a product launch — it reads more like a strategic brief aimed at CIOs and heads of engineering who are trying to figure out where they should be putting their bets.
The context matters here. OpenAI has spent the last 18 months aggressively expanding its enterprise footprint. The ChatGPT Enterprise tier, launched in mid-2023, was followed by OpenAI for Teams, deeper API integrations, and more recently the rollout of Codex as a serious agentic coding platform. (We covered that shift in depth in our piece on what OpenAI Codex actually does beyond chat.) The B2B Signals report is essentially OpenAI saying: here’s the data on what’s working, here’s what the best customers are doing, and here’s where this is heading.
It’s also, not coincidentally, a competitive move. With Google Cloud pushing Gemini hard into enterprise deals and Anthropic’s Claude gaining serious traction in legal, financial services, and research workflows, OpenAI has every reason to publish evidence that its customers are getting measurable results.
What the Data Actually Shows
The report identifies a clear tier structure in enterprise AI adoption. OpenAI calls the top cohort “frontier firms” — and the defining characteristic isn’t budget or company size. It’s depth of integration.
Here’s what separates frontier firms from the rest, according to the research:
- They’ve moved beyond prompt-and-response workflows. Instead of employees asking ChatGPT questions and copying answers into documents, frontier firms are running multi-step agentic pipelines where AI takes actions, checks its own work, and hands off outputs to other systems.
- Codex is central to their engineering stack. The report highlights Codex-powered workflows as a common thread — specifically, teams using Codex agents to handle issue triage, code review, test generation, and deployment prep autonomously rather than as a copilot that waits for human prompts.
- They measure AI impact on business outcomes, not usage metrics. The lagging companies track “seats” and “queries per month.” The leading ones track time-to-ship, error rates, and engineer throughput.
- They’ve assigned dedicated AI ownership internally. Not just a Chief AI Officer title, but actual cross-functional teams with authority to change processes — not just recommend them.
- They iterate fast on failure. Frontier firms treat broken agentic workflows the way good engineering teams treat production bugs: with post-mortems, root cause analysis, and rapid fixes.
The report also points to a compounding dynamic that’s hard to overstate. Each agentic workflow that gets deployed generates data about how AI performs in that specific context. Frontier firms use that data to fine-tune behavior, build institutional knowledge into their systems, and raise the baseline for what “normal” looks like in their operations. Companies that haven’t started that loop yet are falling further behind with each passing quarter — not because the technology is unavailable to them, but because the organizational learning curve takes time to climb.
The Codex Agentic Workflow Shift
Codex deserves particular attention here because the B2B Signals report frames it as the primary vehicle for agentic transformation in technical teams. If you’ve been following our coverage, you’ll know that OpenAI repositioned Codex earlier this year not as a code completion tool but as an agent capable of working directly with issue trackers and software systems. The B2B data appears to validate that positioning.
In practice, what frontier firms are doing with Codex looks something like this: a developer opens a ticket, Codex agents read the issue context, pull relevant code, propose and implement a fix, run tests, flag conflicts, and surface a PR for human review. The human’s job shifts from writing code to reviewing decisions. That’s not a marginal efficiency gain — it’s a different model of how engineering teams function.
What “Durable Competitive Advantage” Actually Means
OpenAI uses the phrase “durable competitive advantage” throughout the report, and it’s worth unpacking what that means in practice — because it’s not just about being faster than a competitor today.
The argument the report makes, implicitly, is that AI-transformed workflows create switching costs. Once a company has built agentic pipelines deeply tied to its internal data, its codebase, its documentation, and its processes, those pipelines become harder to replicate than any single AI model capability. The model itself can be swapped out. The institutional integration can’t be easily copied.
This is actually similar to the argument that made enterprise software sticky for decades. SAP didn’t win because it had the best software — it won because once your entire supply chain is running through it, you don’t leave. OpenAI is betting that deep agentic integration creates the same kind of gravity.
Who This Report Is Really Written For — And Who Should Be Worried
On one level, B2B Signals is market research dressed up as thought leadership. It’s designed to give enterprise buyers permission to spend more, move faster, and push for deeper integration. That’s fine — it’s honest about what it is.
But the competitive implications for mid-market companies are genuinely uncomfortable. The report implicitly frames this as a window that won’t stay open forever. If the compounding loop of agentic deployment → data collection → workflow improvement is real, then organizations that delay meaningful AI integration aren’t just missing short-term productivity gains — they’re watching competitors build structural advantages that get harder to close over time.
I wouldn’t be surprised if the next 12 months see a sharp acceleration in enterprise AI consultancy deals, specifically around agentic workflow design. There’s a whole services industry about to emerge around helping companies catch up to frontier firms — similar to what happened with cloud migration between 2015 and 2020.
For context on how OpenAI is positioning itself broadly in the enterprise market, it’s worth reading our analysis of the OpenAI-AWS partnership, which gives these agentic capabilities a much wider distribution path than OpenAI’s direct sales channel alone.
Where Does This Leave Google and Anthropic?
Google’s enterprise AI push through Gemini for Google Workspace and Vertex AI is strong in organizations already deep in the Google stack. But the B2B Signals framing — with its emphasis on Codex-driven engineering workflows — is clearly aimed at a segment where Google hasn’t yet established the same depth. Anthropic’s Claude 3.5 and beyond has been gaining real ground in knowledge-work-heavy enterprises, particularly finance and law, but it doesn’t have the equivalent of Codex for agentic software development at scale. At least not yet.
OpenAI’s timing here feels deliberate. Publishing a report that centers Codex as the mechanism for enterprise transformation, while simultaneously rolling out GPT-5.5 as its flagship work-oriented model, creates a coherent narrative: the best model, the best coding agent, and now the research to show it’s working. Whether that narrative holds up over the next product cycle depends on execution — but as a strategic communication move, it’s well-constructed.
Key Takeaways From OpenAI’s B2B Signals Report
- Frontier enterprises are running agentic, multi-step AI workflows — not just using AI as a chat interface
- Codex-powered pipelines are driving measurable engineering productivity gains in leading organizations
- The gap between high-adoption and low-adoption enterprises is compounding over time, not staying static
- Competitive advantage comes from deep integration with internal systems and data, not from access to models alone
- Internal AI ownership — with real authority to change processes — is a consistent trait in frontier firms
- OpenAI’s enterprise strategy is now clearly built around agentic workflows as the core value proposition, not model capabilities in isolation
Frequently Asked Questions
What is OpenAI’s B2B Signals report?
It’s a research publication from OpenAI that analyzes how its top enterprise customers are adopting and scaling AI — specifically agentic workflows powered by tools like Codex. The report is based on real usage data and enterprise interviews, and is aimed at helping business leaders understand what separates high-impact AI adoption from more superficial deployments.
What are “frontier firms” according to the report?
OpenAI uses the term to describe enterprises that have moved beyond using AI as a productivity add-on and are instead restructuring workflows around agentic AI systems. These firms are characterized by deep integration, dedicated internal AI ownership, and a focus on business outcome metrics rather than usage statistics.
How does Codex fit into the enterprise AI strategy described in the report?
Codex is positioned as the primary vehicle for agentic transformation in engineering teams. The report highlights how frontier firms use Codex agents to autonomously handle tasks like issue triage, test generation, and code review — shifting engineers from writing code to reviewing AI-generated decisions.
Should companies that haven’t adopted agentic AI yet be worried?
The report argues that AI-transformed workflows create compounding advantages over time, meaning the gap between early adopters and laggards widens with each quarter. It’s a credible concern, though the degree of urgency depends heavily on the competitive dynamics in each specific industry.
The honest read on B2B Signals is that it’s part research, part recruitment tool — OpenAI wants more enterprises committing to deeper integration, and publishing proof points from the companies already doing it is a smart way to make that case. Whether that push toward deeper enterprise entrenchment turns out to be good for customers in the long run is a question worth watching closely as agentic AI moves from pilot programs into the operational core of major organizations.
