Anthropic did something unusual: it surveyed 81,000 real users and asked them what they found most valuable after using Claude.
The answer surprised many.
Only 40% said they felt faster or saved time.
48% said they "learned skills they didn't have before" and "could do things they couldn't do before."
The biggest gain isn't speed — it's capability
This result is worth pausing over.
How do we usually describe the value of AI tools? "How many hours saved," "how many times more efficient" — the speed framework.
But among these 81,000 users, nearly half said the biggest benefit AI brought them wasn't speed, but expanding what they could do.
A few examples:
- A product manager who can't code can now build prototypes themselves
- An entrepreneur without a legal background can now understand contract terms on their own
- An engineer with no design background can now produce decent UI
The value of this kind of "new skill emergence" simply can't be calculated in terms of "saving X hours."
On average, users gave Claude a productivity improvement rating of 5.1 out of 7. Not perfect, but this is self-reported data from real users — some gave 1, some gave 7, and the average came to 5.1.
Who benefits most, and who worries most
The two groups that benefited the most:
One is management roles and math/computer-related professions. They used Claude to expand the boundaries of their professional capabilities, doing things that previously required outsourcing or hiring experts.
The other group is somewhat unexpected: low-income workers. They are using Claude to open up additional income streams — side hustles, freelance gigs, projects they couldn't do on their own before.
These two groups look very different on the surface, but the logic behind them is consistent: AI helped them do things they couldn't do before.
The most anxious group:
1 in 5 expressed concern about losing their jobs. Early-career workers showed significantly higher anxiety levels than experienced professionals.
Not surprising. Entry-level jobs are more directly affected, and the barriers of seniority and judgment haven't been built yet.
The strange dilemma of creators
The most interesting finding was among creators — visual artists and writers.
They showed a U-shaped distribution: some said AI slowed them down, that the tools were too rigid and unsuitable for creative workflows; others said it greatly improved their efficiency.
But regardless of which end they were on, their fear of being replaced was above average.
Anthropic's report used one phrase: visual artists and writers found AI "too rigid and limiting."
This feedback illustrates one thing: AI's improvement in execution-type tasks is real, but in scenarios requiring a high degree of personal style and creative judgment, current models still have clear shortcomings.
In short: AI helps with "delivery," but it's still far from helping with "creation."
A limitation of the data
81,000 people sounds like a lot. But these are individual users who voluntarily participated in the survey on Claude.ai, and do not include enterprise deployments.
In enterprise settings, Claude is typically used in a more standardized and process-driven way — in that scenario, the value of "getting faster" may be more prominent, and the proportion of "learning new skills" may be lower.
So what this survey tells us is: the greatest value perceived by individual users is skill expansion, not just speed improvement.
The enterprise side may be a different story.
But regardless of what enterprise data shows, the answers from these 81,000 people make one thing clear:
The value of AI tools has been underestimated by the "time savings" framework.
What's really happening is that a group of people suddenly find they "can do things they couldn't do before."
This kind of expansion is harder to replace — and more interesting — than being ten minutes faster.
Source: CocoLoop, Claude survey: new capabilities beat speed as top AI benefit, but creatives feel left behind (The Decoder)