OpenAI’s first State of Enterprise AI report, published in December 2025, puts hard numbers on something many UK business leaders have suspected: the organisations getting serious about AI are pulling away from everyone else. The data, drawn from over 1 million business customers and a survey of 9,000 workers across nearly 100 enterprises, tells a story of rapid acceleration at the top and stagnation in the middle.
The headline figure that should concern every board: frontier workers (the top 5% by usage) send 17 times more coding messages than the median employee. That’s not a small edge. It’s a different way of working entirely.
What the numbers actually show
Before getting to the strategic implications, it’s worth sitting with the data. Some of these figures are genuinely striking.
| Metric | Figure | Context |
|---|---|---|
| ChatGPT Enterprise message growth | 8x year-over-year | Aggregate weekly messages |
| API reasoning token consumption | 320x per organisation | Year-over-year increase |
| Custom GPT/Project weekly users | 19x year-to-date increase | 20% of all Enterprise messages now go through GPTs |
| Median sector growth | 6x year-over-year | Technology sector led at 11x |
| Frontier vs median worker messages | 6x gap | 95th percentile vs median |
| Frontier vs median coding messages | 17x gap | Largest category gap |
| Time saved (general workers) | 40-60 minutes per day | Self-reported via survey |
Critical context: These numbers come from OpenAI’s own customer base — organisations already paying for Enterprise seats. The gap between these firms and organisations not yet using enterprise AI tools is likely far larger than anything measured here.
The 320x increase in API reasoning token consumption per organisation deserves particular attention. It suggests that companies aren’t just chatting with AI more — they’re building it into production systems. When reasoning tokens scale that dramatically, it means models are being embedded in products and workflows, not just used for ad-hoc queries.
The real story: a workforce splitting in two
The most important finding in this report isn’t about adoption rates or time saved. It’s about the growing divide between workers who have figured out how to use AI deeply and those who haven’t.
Consider the task-level data:
- Coding: frontier workers send 17x more messages than median
- Writing and communication: 11x gap
- Analysis and calculations: 10x gap
- Information gathering: 9x gap
- How-to guidance: 9x gap
These gaps exist within the same organisations, using the same tools, with the same access. The constraint isn’t technology. It’s behaviour.
Strategic insight: Workers who use AI across roughly seven task types report five times more time saved than those using it for about four tasks. The returns don’t come from doing one thing well with AI — they come from breadth of application.
There’s a compounding effect at work. Workers who save more than 10 hours per week aren’t just using AI more frequently. They’re using multiple models, engaging with more tools (reasoning, data analysis, search), and applying AI across a wider range of tasks. The 10+ hour savers consume 8x more intelligence credits than those reporting zero time savings.
Meanwhile, even among monthly active Enterprise users, 19% have never tried data analysis, 14% have never used reasoning, and 12% have never used search. These are paid features sitting unused.
What’s happening at the firm level
The individual gaps mirror what’s happening at the organisational level. Frontier firms (top 5% by adoption intensity) generate approximately 2x more messages per seat than the median enterprise. For GPT messages specifically, it’s 7x.
What separates these firms isn’t budget or access. The report identifies five consistent practices:
1. Deep system integration. Leading firms connect AI to their internal data through connectors and APIs. One in four enterprises still hasn’t taken this step — their AI operates blind to company context.
2. Workflow standardisation. They build and share reusable GPTs and API-powered assistants. BBVA, for example, regularly uses more than 4,000 custom GPTs. That’s not experimentation. That’s institutional embedding.
3. Executive sponsorship. Clear mandates from leadership, dedicated resources, and space for experimentation. Without this, adoption stays grassroots and uneven.
4. Data readiness. They codify institutional knowledge into machine-readable formats, build APIs for data pipelines, and run continuous evaluations against real-world outcomes.
5. Change management infrastructure. Centralised governance combined with distributed enablement through embedded AI champions across teams.
Reality check: OpenAI releases a new feature or capability roughly every three days. The primary constraint for most organisations is no longer model performance or tooling — it’s organisational readiness to absorb and deploy these capabilities.
The financial case is building
A 2025 Boston Consulting Group study found that over three years, AI leaders achieved:
- 1.7x revenue growth compared to laggards
- 3.6x greater total shareholder return
- 1.6x EBIT margins
| Metric | AI leaders vs laggards (3-year) |
|---|---|
| Revenue growth | 1.7x |
| Total shareholder return | 3.6x |
| EBIT margins | 1.6x |
These correlations don’t prove causation, and the BCG study has limitations. But when combined with OpenAI’s case study evidence, a pattern forms. Intercom reports a 48% latency reduction and 53% call resolution rate using AI voice agents. Lowe’s sees conversion rates more than double when customers engage with their AI assistant. Indeed found that job seekers using their AI career tool are 38% more likely to be hired.
Strategic reality: The question for UK boards is shifting from “should we invest in AI?” to “how far behind are we, and how quickly can we close the gap?”
What UK organisations should actually do
The report’s findings, while drawn from a global (and heavily US-weighted) customer base, carry direct implications for UK enterprises. The UK and Germany rank among the largest ChatGPT Enterprise markets outside the US by customer count, and international API customer growth has exceeded 70% over six months.
For organisations just starting
Priority one: connect AI to your data. One in four enterprises hasn’t turned on data connectors. Without organisational context, AI tools operate at a fraction of their potential. This is the single highest-return action for early-stage adopters.
Priority two: identify your frontier workers. They already exist in your organisation. Find the people saving hours per week and learn what they’re doing differently. Their workflows become your templates.
Priority three: expand task coverage, not just frequency. The data shows returns compound with breadth of use. Don’t optimise for one AI use case — encourage exploration across writing, analysis, coding, research, and creative work.
For organisations already deploying
Build the GPT library. Frontier firms standardise and share reusable AI workflows. Every team should be creating, curating, and publishing custom GPTs for their domain. Make discovery easy.
Invest in non-technical AI literacy. Coding-related messages from non-engineering teams grew 36% over six months. AI is blurring traditional role boundaries. Organisations that train non-technical staff to use AI for data analysis and automation will see disproportionate returns.
Run continuous evaluations. Leading firms don’t just deploy AI — they measure its performance against real-world outcomes and iterate. Build feedback loops between AI outputs and business metrics.
Take action: Start with an audit. How many of your Enterprise seats have never used reasoning, data analysis, or search features? That unused capability represents your largest immediate opportunity.
Four challenges nobody’s talking about
1. The measurement problem. Self-reported time savings of 40-60 minutes per day are encouraging but inherently unreliable. Organisations need objective productivity metrics, not survey responses. The firms that build proper measurement systems will make better investment decisions.
2. The shadow gap. The report measures active Enterprise users but says nothing about the majority of the workforce without Enterprise access. In most organisations, the 6x frontier-to-median gap visible in the data understates the actual disparity across the entire workforce.
3. The vendor lock-in question. With 9,000 organisations processing over 10 billion tokens and nearly 200 exceeding 1 trillion, deep API integration creates significant switching costs. UK organisations should think carefully about multi-vendor strategies, especially given EU AI Act compliance requirements that may favour architectural flexibility.
4. The skills reshaping risk. 75% of users report completing tasks they previously couldn’t perform. That’s powerful for individual productivity, but it also means job descriptions, performance standards, and team structures may need fundamental redesign. Few organisations are planning for this.
Hidden cost: The report frames the frontier-to-laggard gap as an opportunity. It is. But it’s also a risk. Organisations that fall too far behind may find the gap becomes structurally difficult to close as leading firms embed AI into their competitive advantage.
What this means for your strategy
The core message from OpenAI’s report is straightforward: enterprise AI has moved past the experimentation phase. The organisations generating measurable value aren’t doing anything magical — they’re being systematic about integration, standardisation, and change management.
Three factors will determine which UK organisations capture the most value:
-
Breadth over depth. Push AI adoption across task types, not just within a few workflows. The compounding returns favour organisations where AI is a general-purpose tool, not a specialist one.
-
Organisational readiness over model capability. The technology is already more capable than most firms have embedded into their workflows. The constraint is organisational, not technical.
-
Measurement rigour. Move beyond “hours saved” surveys. Build evaluation frameworks that connect AI usage to business outcomes — revenue, customer satisfaction, cycle times, error rates.
Your next steps
- Audit current AI tool utilisation across your Enterprise seats — identify unused features and low-adoption teams
- Map your frontier workers and document their workflows for wider distribution
- Connect AI tools to internal data systems if you haven’t already
- Build a shared GPT/workflow library with clear discovery mechanisms
- Establish baseline metrics for AI-influenced business outcomes
- Review your organisational readiness against the five practices identified in the report
Source: OpenAI — The State of Enterprise AI (December 2025)
This analysis was produced by Resultsense to help UK organisations make practical sense of enterprise AI developments. The insights and recommendations reflect our independent assessment of the report’s findings and their implications for UK businesses.