The “6x productivity gap” headline is based on a real OpenAI finding, but it overstates what was measured. In The State of Enterprise AI: 2025 Report, OpenAI says workers at the 95th percentile of adoption intensity sent six times as many ChatGPT messages as the median worker. That is a usage gap, not evidence that they produced six times as much valuable work, earned six times as much revenue, or saved six times as many hours.
The report does associate broader, more integrated AI use with greater self-reported time savings. The useful lesson for employers and employees is therefore about workflow adoption—not a guaranteed sixfold return from buying an AI subscription.
What OpenAI actually measured
OpenAI’s report combines de-identified, aggregated usage data from enterprise customers with a survey of approximately 9,000 workers across almost 100 enterprises. OpenAI describes the data as covering its enterprise customer base, which at the time included more than one million business customers and more than seven million ChatGPT workplace seats. Those are company- and seat-level figures; they do not mean every customer contributed equally to the worker analysis.
The report defines “frontier workers” as people at the 95th percentile of adoption intensity. They sent six times as many ChatGPT messages as the median worker. The median is the midpoint of the distribution, not the average and not a group made up entirely of non-users.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
| Finding | What it measures | What it does not establish |
|---|---|---|
| 6x overall gap | Messages sent by 95th-percentile users versus the median worker | Six times the output, income, or economic value |
| 16x data-analysis gap | Use of the data-analysis tool by frontier data-analysis users versus median users | Sixteen times more accurate or valuable analysis |
| 17x coding gap | Coding-related messages from frontier workers versus the median | Seventeen times more software shipped |
OpenAI’s primary report and its full PDF provide the definitions and comparisons: report page and full report PDF.
The distinction matters because message counts are ambiguous. More messages could reflect useful iteration, complex work, automation, or a multi-step workflow. They could also reflect repeated prompting, poor answers, low-value experimentation, or extra correction work. Usage intensity is best treated as a leading indicator of adoption maturity, not a productivity score.
What the report says about time savings
OpenAI reports several encouraging associations, but most are survey responses or relationships in usage data rather than controlled productivity measurements.
Rank #2
- 75% of surveyed enterprise users said AI improved the speed or quality of their output.
- ChatGPT Enterprise users attributed 40–60 minutes saved per active day to AI use.
- Data science, engineering, and communications workers reported approximately 60–80 minutes per day saved.
- Workers using AI across roughly seven task types reported about five times more time saved than those using it across roughly four task types.
- People reporting more than 10 hours of weekly savings tended to use more tools, models, and task categories.
These figures should be written as “reported saving” or “was associated with,” not as independently verified hours or proof that AI caused the result. OpenAI is both the vendor and publisher of the analysis, and the report does not present a randomized control group or an audited measure of output, quality, revenue, or profit.
Is the comparison power users versus non-users?
No. The central comparison is the top 5% of adoption-intensity users against the median worker. The median worker may already use ChatGPT regularly. Saying “power users versus everyone else” suggests a much wider divide than the data actually tests.
Among monthly active enterprise users, OpenAI says 19% had never used data analysis, 14% had never used reasoning, and 12% had never used search. Among daily active users, those shares fell to 3%, 1%, and 1%, respectively. The pattern supports a link between frequency and breadth of use; it does not show that activating every feature automatically improves performance.
What frontier users do differently
They cover more kinds of work
Frontier users apply AI across research, drafting, analysis, coding, communications, support, and other task categories instead of reserving it for occasional text generation. Breadth gives them more opportunities to compound small time savings.
They use advanced tools
Reasoning, data analysis, search, image generation, and coding tools appear more often in intensive usage. Technical work is not confined to traditional engineering teams; analysts and other non-engineering employees may use the same capabilities when their workflows support it.
They build repeatable workflows
The practical difference is usually not clever prompt wording. It is reusable context, templates, custom assistants, connectors, staged tasks, and human approval checkpoints. OpenAI’s report says frontier firms generated about twice as many messages per seat as the median enterprise and about seven times as many messages to GPTs, suggesting greater use of customized or organization-specific workflows.
They delegate parts of multi-step work
Instead of asking for an isolated answer, an experienced user may have the system retrieve information, analyze it, draft an output, and prepare a review checklist. Humans still decide, verify, and approve.
The company-level divide
OpenAI says weekly enterprise message volume grew approximately eightfold in aggregate since November 2024, while the average worker sent about 30% more messages over that period. It also reports that the median industry grew by more than sixfold over the prior 12 months, with technology growing approximately 11-fold.
Those statistics indicate expanding use and integration, not automatically better financial performance. A firm can generate more messages while adding review costs, duplicating work, or encouraging low-value experimentation. The report cites examples involving revenue growth, customer experience, shorter product-development cycles, coding, analysis, and faster document production, but those are case examples rather than a randomized estimate of average return on investment.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchBest Value
What the sixfold number cannot prove
- That frontier workers complete six times as much work.
- That they save six times as many hours.
- That AI increased company productivity sixfold.
- That the median worker is wasting 83% of an available AI benefit.
- That a more expensive plan will produce a sixfold return.
Frontier users may already be more skilled, more senior, better supported by data infrastructure, assigned more AI-suitable work, or more willing to experiment. Heavy use can also signal difficult, iterative work rather than superior efficiency. Productivity may appear as greater scope or quality instead of fewer hours, while some effort shifts into fact-checking, editing, security review, and exception handling.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How a company can close the useful gap
- Choose measurable workflows. Start with recurring research, document review, data cleaning, reporting, support triage, coding and test generation, or spreadsheet production.
- Set a baseline. Record cycle time, time to first draft, approval time, error rates, rework, quality scores, and customer response or resolution time before changing the process.
- Train for task selection and verification. Employees need to know when AI is appropriate, how to provide context, how to stage a task, and how to verify calculations, citations, and claims.
- Provide approved context. Connect authorized knowledge sources and define permissions, retention, audit, and human sign-off rules.
- Create reusable components. Templates, custom assistants, standard operating procedures, and connectors make good use repeatable instead of dependent on a few enthusiasts.
- Subtract supervision costs. Include data preparation, editing, fact-checking, security review, failed attempts, training, and change-management time in the business case.
- Scale only when outcomes improve. Message volume can diagnose adoption, but cycle time, quality, rework, and business results should determine whether a workflow stays.
Buying implications for individuals and businesses
A subscription does not create the organizational context associated with frontier-firm adoption. Plan choice should follow the workflow and governance requirement.
| Option | Typical fit | Important qualification |
|---|---|---|
| ChatGPT Plus | Individual testing of research, reasoning, analysis, coding, and personal workflows | Listed at $20 per month on OpenAI’s consumer pricing page; no shared enterprise administration by itself |
| ChatGPT Pro | Individuals needing higher access for intensive personal use | Listed at $200 per month; price does not guarantee business value |
| ChatGPT Business | Small and midsize teams needing shared workspace, connectors, centralized billing, and baseline governance | OpenAI’s page lists $20 per user per month annually or $25 monthly, with a two-seat minimum; pricing checked August 16–18, 2026 |
| ChatGPT Enterprise | Large organizations needing identity, retention, support, residency, procurement, and formal controls | Custom pricing; contact sales |
| API workflow | AI embedded in internal applications, ticketing, document, or approval systems | Economics vary with model, context, tool calls, retries, and human review |
OpenAI’s business details are at its Business pricing page and Business Help Center article. The consumer comparison is at ChatGPT pricing. Business data-use, retention, residency, and feature availability can vary by plan and geography, so buyers should verify current terms.
Organizations comparing alternatives such as Claude should run the same representative tasks through each system and score quality, review time, security controls, connectors, administration, and total cost. Claude’s current pricing page is claude.com/pricing. For custom applications, OpenAI’s platform entry point is platform.openai.com.
Recommended Free Tools
The accurate takeaway
The popular headline, including the framing used by VentureBeat, compresses a real statistic into a misleading conclusion. OpenAI found a sixfold difference in ChatGPT message volume between frontier users and the median worker. Its survey also found that broader use was associated with greater reported time savings.
The durable business lesson is narrower and more useful: the largest divide is between occasional chatbot use and AI embedded in repeatable, multi-step work. Companies should measure that integration through output quality, cycle time, rework, and business outcomes—not treat a rising message counter as proof of productivity.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
