OpenAI enterprise AI report: Why adoption stalls

OpenAI recently released their State of Enterprise AI 2025 report, and the data confirms what we've been saying: AI can do far more than most businesses realize. But buried in the report is a paradox that reveals why enterprise AI adoption has stalled—and why OpenAI's own product strategy is part of the problem.

The report details an uneven distribution of results within companies using OpenAI, skewed towards software development use cases. 73% of engineers report faster code delivery, while Codex, OpenAI’s coding tool, has doubled in usage in 6 weeks. For technical teams, ChatGPT Enterprise is delivering.

AI is enabling capability expansion

But what about non-technical teams?

The report shows that business teams, such as accounting, finance, analytics, and communications also report large productivity benefits, but in an unexpected way:

"Among ChatGPT Enterprise users, coding-related messages have increased across all functions, and outside of engineering, IT, and research, coding-related messages have grown by an average of 36% over the past six months."

Dane Vehey of OpenAI tags this as capability expansion.

Business users are getting value from AI, but primarily when they use it for technical tasks like coding. However, the advanced, non-coding capabilities of AI remain underutilized.

The OpenAI report puts it this way:

"Many have not tried some of the most capable tools…19% have never used data analysis, 14% have never used reasoning, and 12% have never used search."

In other words, ChatGPT is helping salespeople become vibe coders, but it's not helping them hit their revenue goals for the quarter.

The capability overhang of enterprise AI

OpenAI acknowledges that there’s a gap between what their current technology can achieve and what most organizations use them for. "Models are capable of far more than most organizations have embedded into workflows, and this presents an opportunity for firms," OpenAI’s report details.

This is what Microsoft's Kevin Scott calls the “capability overhang” - the growing gap between what AI models can actually do and what businesses do with them. The models have leapfrogged ahead, but enterprise adoption outside of developers hasn't caught up.

Part of the problem is caused by OpenAI’s own product experience, which prioritizes consumers and developers, not business users. OpenAI recently released specialized models for coding, and a model update to make the AI more friendly.

Contrast this with OpenAI’s enterprise features, which users report feels bolted on rather than built in.

3 barriers to enterprise adoption of AI

The State of Enterprise AI report is notably silent on the UX barriers preventing business teams from capturing AI value:

Context switching friction. Business projects live in Slack and Microsoft Teams, not in a separate AI chat application. OpenAI’s tools force users to bounce between tools, which encourages work silos instead of cross-team collaboration.

Data connectivity constraints. Enterprise ChatGPT connects to a limited set of data sources. But real business intelligence requires access to CRMs, ERPs, data warehouses where institutional knowledge actually lives. If the AI can't see your business context, it can't reason about your business.

Single-model lock-in. OpenAI's enterprise offering uses OpenAI models exclusively. But different models excel at different tasks. Anthropic's Claude handles nuanced analysis differently than GPT-4. Google's Gemini processes certain document types more effectively. Choosing one vendor means accepting capability tradeoffs.

At Adapt, we build for business use cases so the company can collectively be successful with AI. Our platform connects to the data sources businesses actually use, works inside Slack where teams collaborate, and lets you choose the model for each job. When a sales leader asks about pipeline trends, they shouldn't need to export data, switch applications, or understand token limits.

Read more about capability overhang and how your business can gain the most from AI in our report, “AI for Startup Leaders”

FAQ

Why is AI delivering value for developers but not for business teams?

AI tools are optimized for technical workflows where outputs can be immediately tested, validated, and shipped. Business teams, by contrast, are given the same interfaces without the context, data access, or workflow integration needed to turn AI output into real outcomes.

What does “capability overhang” mean?

Capability overhang describes the widening gap between what modern AI models are technically capable of and how organizations actually use them. The technology has advanced faster than the products, workflows, and incentives that would let non-technical teams benefit from it.

Why are non-technical teams primarily using AI for coding?

Coding tasks provide clear prompts, fast feedback, and obvious success criteria, even for non-engineers. Core business work—forecasting, decision-making, coordination—requires shared context and trusted data that today’s AI tools rarely have access to.

What barriers prevent AI from fitting into business workflows?

AI tools force users to leave the systems where work already happens and operate with partial or disconnected data. Locking teams into a single model further limits their ability to match the right AI capability to the task at hand.

What would an AI platform look like if it were designed for business teams?

Adapt is built to meet teams where work already happens, operating inside Slack and connecting directly to the systems that hold business context. Adapt helps teams move from insight to action without friction.