Anthropic published a new study on June 26, 2026, and the headline finding cuts against almost everything the average leader assumes about AI and work. The people who hand the most work to Claude, including the most fully automated work, are the ones who feel best about their own pay, their job security, and the value of their skills next year.
That deserves a second of your attention, because the fear most professionals carry runs the opposite direction. The story in everyone's head is that the heaviest AI users are automating themselves out of a paycheck. The data says the heaviest users are the most confident they will be fine.
The report is called "Cadences," part of Anthropic's ongoing Economic Index. The survey launched in April 2026, and the final linked sample covers about 9,700 people whose survey answers were tied to their actual Claude usage. That linkage is what makes this different from a poll. It compares what people say about their future against what they do in the product.
The paradox in one sentence
Here is the core result, in Anthropic's words:
"Across all six dimensions, people with a higher share of automated sessions feel more optimistic about the effect of AI on their job outcomes next year compared to those who use Claude more augmentatively. We saw the largest effects on expectations about positive impacts on future pay and ability to find a job."
Read that again. More automation, more optimism. The largest effects showed up on the two things people worry about most: future pay and the ability to find a job.
The report puts a finer point on it:
"the people who delegate to Claude the most are the most optimistic about their future labor market outcomes, and feel their skills are growing in value."
So the professionals delegating the heaviest load to AI are not bracing for replacement. They report their own value climbing.
Why this should change how you think about risk
If you run a team or a P&L, you have probably been treating AI exposure as a threat to manage. This study suggests the opposite read for the people closest to the tool. The threat lives in distance from the work, not proximity to it.
Look at where the worry actually concentrates. Only 10% of respondents rated losing their own job as likely or very likely. But more than one-third said a junior colleague faces a greater-than-60% chance of losing their job in the next year:
"over one third stating that the probability of a junior colleague losing their job in the next year was over 60%."
of respondents rated losing their own job as likely or very likely, while more than a third put a junior colleague's odds of job loss next year above 60%.
Source: Anthropic Economic Index, 2026People feel secure about themselves and nervous about the entry-level seat next to them. That split is the whole story. The confident-for-self, worried-for-others pattern tells you something real about where capability protects a career and where it does not.
The same crowd is not in denial about how much AI will do, either. Over a third of respondents expect AI to handle most or nearly all of their work tasks next year. They see the automation coming, they are running toward it, and they feel safer for it. The thing that turns that combination from scary to encouraging is whether the person is the one directing the work or the one being directed around.
The productivity numbers behind the optimism
The optimism is not vague. It sits on top of concrete gains people report from working with AI:
- 86% report gains in the speed of their work
- 82% report gains in the scope of what they can take on
- 69% report gains in the quality of their output
"large majorities of people report productivity gains in speed, scope, and quality of their work (86%, 82%, and 69%, respectively)."
And the skill story holds up. 57% say AI has made their skills more valuable, and 68% say they are learning more with AI. These are people getting better, not people being hollowed out. The ones delegating the most are the ones reporting the steepest growth.
For a leader, that recasts the spend question. The obvious return is faster output. The durable one is a workforce that reports its own capability rising, the asset that shows up on every future project after this one.
Where this connects to building an AI-capable team
The useful read for anyone who feels behind is this: the people pulling ahead are not the ones avoiding AI to protect their role. They are the ones handing it the most work and watching their own footing get stronger. Avoidance is the riskier position, not the safe one.
That is the spine of The 7 Levels of AI Proficiency, the framework we use at LaunchReady to describe how a person grows from first contact with these tools all the way to directing systems of them. The early levels are about getting comfortable and writing good instructions. The higher levels are about designing the work itself, then managing AI to carry it while the human stays in the loop on judgment, quality, and direction. The Anthropic data, drawn from 9,700 people's real behavior, points the same direction it does. The professionals who climb toward managing the work AI does report feeling more secure, not less.
For the junior-colleague worry, this is the practical answer. The entry-level seat looks exposed when the only thing it offers is task execution, because task execution is exactly what AI now does at scale. The same seat looks durable when the person in it is learning to design and direct the work. The question for your team is whether you are teaching junior people to manage AI or leaving them to compete with it.
A second finding worth a leader's attention
"Cadences" earns its name from a methodology change. Anthropic shifted from sampling usage every seven days to sampling it hourly, so the report can see rhythms it could not see before. The patterns are a reminder that this technology is already woven into daily life, not held at the office.
Personal, non-work use of Claude rises from around 35% on weekdays to just under 50% on weekends. Specific tasks spike at specific hours. Recipe requests run 2.3 times more frequent at 6 p.m. than average. Sleep advice peaks around 5 a.m. Tax questions surged around the US filing deadline.
The strategic point underneath the trivia: your people are already fluent with AI in the parts of their lives where no one is grading them. The comfort is there. The work is connecting that comfort to how the business actually runs.
One more pattern deserves a flag for anyone thinking about who gets ahead. Women made up only 12% of the linked respondent sample, and their share of automated sessions ran 0.33 standard deviations lower than men's, about 7.3 percentage points. If the most automated users are the most optimistic about their futures, then uneven access to that habit becomes an uneven distribution of confidence and capability. For a leader building a team, that is a distribution to watch and to even out on purpose.
Here is a concrete next step. Pick one person on your team who still keeps AI at arm's length, and one piece of recurring work they own. This week, sit with them and map that work into steps an AI tool could carry while they direct and check it. You are not testing the tool. You are starting the climb that the most confident people in this study have already begun.
Related reading: Level 5: The Captain (Design Thinker).
Sources
Frequently Asked Questions
Does this study say AI will not cost jobs?
No. Over a third of respondents expect AI to do most or nearly all of their tasks next year, and more than a third believe a junior colleague has a better-than-60% chance of losing their job in the coming year. The finding is narrower and more useful: among the people surveyed, heavier AI use lined up with more optimism about their own pay and prospects, not less.
Is this just optimism from people who happen to like the tool?
The sample links survey answers to real usage data on about 9,700 people, so it compares stated expectations against measured behavior rather than relying on attitudes alone. That does not prove causation, but it is a stronger basis than a standalone poll.
What is the one thing to take from this as a manager?
Proximity to the work, not distance from it, is where confidence and skill growth showed up. Putting your people in position to design and manage what AI does looks more protective than keeping them away from it.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.