Output Used to Prove You Cared and AI Just Made It Meaningless

For most of working life, doing a lot was how you proved you cared. Fast Company made the case that this no longer holds, arguing that the most important skills in today's workplace are judgment and discretion because the cost of output has never been cheaper. If you have noticed that working hard is not landing the way it used to, that is the reason. The signal you were sending has been devalued, not your effort. At Metaintro, we track this because it quietly changes what a good performance review looks like, and most people have not been told.

Why did volume ever count as proof of effort?

Because it was costly to produce, and costly signals are the only ones that work. Productivity at work used to be measured largely by output, and while results and impact were always the flagship markers, you could get a long way by producing a high volume of just fine output. That worked because output used to be expensive. You had to think and devote energy to what you were creating, and the effort itself signalled that you cared enough to think it through.

The consequence is that visibly caring about your job was typically enough to keep you from being fired. Notice how much of workplace culture that one sentence explains. Long documents, thorough decks, fast replies, visible activity. None of those were ever the point, but all of them were evidence. Most of us are hard wired to believe output equals productive and productive equals good, and that belief was reasonable right up until the cost structure underneath it changed. The same repricing is now making human judgment the scarce input across almost every desk job.

What actually broke the signal?

Speed. The mental model breaks in a world where you can paste a colleague's proposal into an AI assistant, say that you would like a different direction, and get 1,000 words of modestly well reasoned prose. The language is coherent, probably more formally correct than most people can write, and it took nine seconds.

Sit with the nine seconds, because that is the entire argument. If a thousand competent words cost nine seconds, then a thousand competent words tell your manager nothing about your commitment, your understanding or your care. They tell them you have a subscription. Every workplace signal that was built on the old cost of production is now unreliable in the same way, which is a genuinely disorienting thing to happen to a set of habits most people built over a decade or more. Junior workers feel it first, which is why the entry level squeeze arrived before anyone changed the review template.

Are employers still rewarding the wrong thing?

Some are, visibly, and the scale of it is now documented. The Walt Disney Company has given thousands of its product and technology employees access to an internal AI Adoption Dashboard that tracks their use of Claude and Cursor, with request counts, token consumption and a leaderboard of top users, according to Business Insider reporting summarised by HR Executive. Employees responded exactly as you would expect. They started gaming it, a practice now called tokenmaxxing.

The numbers are the argument. One Disney employee invoked Claude about 460,000 times over nine work days in mid-April, roughly 51,000 times a day, and one case involved automated agents accumulating 234.2 million tokens. Nobody made 460,000 considered decisions in nine days. What was measured was activity, and activity is exactly what has become free.

Disney is not alone. Meta ran an internal dashboard showing more than 60 trillion tokens consumed over 30 days before shutting it down over external sharing concerns, Visa reported 1.9 trillion tokens consumed monthly as of March and began rewarding teams for productive use, and JPMorgan Chase has used internal dashboards to monitor adoption. Amazon sits in the same group of large employers pushing AI adoption through visible measurement. The problem is now large enough to have attracted governance. The Linux Foundation has said it intends to launch the Tokenomics Foundation, a vendor neutral programme to develop benchmarks and best practices for managing token based AI spending. When an open source umbrella body starts writing standards for your performance metric, the metric is in trouble.

If you work somewhere running a dashboard like this, the practical risk is not that you are being watched. It is that you are being ranked on the one dimension that no longer distinguishes anybody, a pattern already visible where AI usage feeds directly into performance reviews and where monitoring triggered an open backlash at Duolingo.

None of this is an argument against using AI well. Better prompting produces better output, and the people who are good at this are getting meaningfully more done. The point is narrower and more useful. Using the tools well is now the baseline rather than the differentiator, in the same way that touch typing stopped being a competitive advantage once everyone had a keyboard.

Does the wider job market data support this?

It does, and unusually directly. PwC analysed more than one billion job advertisements across 27 countries and territories and found AI increasing the emphasis on human skills such as judgement, creativity and leadership. That is the same conclusion arrived at from job ads rather than from argument.

The structural finding is the useful one for career planning. PwC describes a two track labour market in which professionalised roles, where AI automates routine tasks so human judgement and expertise are emphasised, are seeing twice the growth in available jobs and 42 percent faster salary growth than democratised roles, where AI makes the job itself easier for non experts to perform. Radiologists and recruiters sit in the first group, IT service managers and medical secretaries in the second. The average wage premium for AI skills has risen to 62 percent, and jobs requiring those skills are growing 69 percent against 9 percent for the market overall.

The entry level finding is the one that should change how junior workers spend their time. Analysing 2.4 million entry level jobs in the United States, PwC found AI exposed entry level roles are now seven times more likely to require traditionally senior level human intensive skills such as leadership, creativity or face to face interaction, and openings for those roles grew 35 percent since 2019 while other entry level roles shrank 10 percent. The market has already repriced judgement. Most performance review templates have not.

What does judgment actually look like in practice?

It is easier to recognise than to define, so use the negative test. Judgment is the part of your work that would go wrong if a competent stranger with the same tools did it instead of you. That test is worth applying to your own week, and it is the same test that separates a solopreneur who triples output with AI tools from one who simply triples the volume of work nobody asked for.

It covers deciding what not to do, which almost never appears in a performance review and is often the highest value act of the quarter. It covers knowing when an output is plausible but wrong, which is precisely the failure mode of fast generated work. It covers setting the direction before the work starts, since it is easier than ever to run a million miles per hour in any direction, which means you had better be running in the right one. And it covers the contextual knowledge that never got written down. Research from the Harvard Business School AI Institute describes exactly this category, finding workers hold large amounts of valuable uncodified knowledge such as tacit know how, judgment, communication skills and context specific understanding of customers, projects and processes that is not captured in manuals or company wikis.

Discretion, the second skill named in the original argument, is the quieter half. It is knowing which version of the truth a client needs, when to escalate, what to leave out. None of it generates artefacts, which is why it has always been undervalued, and why it is now the thing worth making visible.

How do you make judgment visible in a review?

Convert decisions into evidence, because judgment leaves no artefact by default. Volume produced its own paper trail, which was part of its appeal. Judgment produces an absence, the problem that never happened, the work correctly not done, and absences do not appear in a self assessment unless you put them there.

Three habits carry the load. Record decisions rather than deliverables, noting what you chose, what you rejected, and what constraint you were working under at the time. Record the redirections, since the moment you noticed something plausible was wrong is the clearest evidence you were not simply passing work through. And record what you declined, because a quarter where you killed two low value projects and shipped one good one is a strong quarter that a volume metric will score as weak.

The framing matters as much as the record. Instead of describing what you produced, describe the decision and its consequence. Getting that framing wrong is one of the mistakes that quietly cost people a promotion, and it lands differently depending on whether your organisation has rebuilt feedback for the generative AI era or still runs a traditional annual cycle.

What should employers measure instead?

The honest answer is that nobody has fully solved this, but the direction is clear enough to act on. If quantity has stopped discriminating between people, the measures that remain have to capture decisions rather than artefacts.

Three shifts are already visible in how thoughtful teams are adapting. The first is measuring the quality of what was rejected rather than the quantity of what shipped, which surfaces the discretion that a volume metric actively punishes. The second is assessing direction setting separately from execution, since the cost of running fast in the wrong direction has risen sharply now that running fast is free. The third is treating tool fluency as a floor rather than a score, because better prompting does produce better output and the people good at it are getting meaningfully more done, which makes it a hiring requirement rather than a ranking dimension.

There is a governance reason to move quickly too. Research from the Harvard Business School AI Institute found that when workers understood their recorded behaviour could train systems that substitute for their expertise, they withheld it, with 41 percent refusing to share past data against 25 percent of those told nothing. A metric that rewards visible activity while quietly harvesting it is a fast route to that outcome, and it compounds when employers demand more output from fewer people.

Employers are already splitting into two camps on what to do with the productivity gain. Some are making AI fluency a condition of advancement, as Accenture has done by requiring AI skills for promotion. Others are testing whether the gain can shorten the week rather than raise the target. Which road an employer takes tells you a great deal about how it will handle the judgement question as well.

What does this mean for your career?

Four moves, roughly in order of urgency.

Stop competing on volume, because you will lose to a subscription. If your visible contribution is quantity of competent output, you are defending a position that costs nine seconds to attack. Move your visible contribution to the decisions.

Find out how your employer actually measures you, and take an AI dashboard seriously as information. Disney's leaderboard, Meta's token counter and Visa's team rewards are all attempts to measure something real with the only number available. If quantity is what is being ranked, you now know the metric is wrong and you know the direction it will eventually move. That is useful lead time, and worth reading alongside how AI native workers are being overseen.

Pick the track deliberately. The PwC split between professionalised and democratised work is the single most actionable finding here, given the difference in job growth and the 42 percent gap in salary growth. Ask honestly which side your role currently sits on, then ask which parts of it could move, keeping in mind that AI is landing hardest on cognitive work rather than on jobs as whole units.

If you manage people, fix the metric before it costs you your best people. Rewarding pure quantity when quantity is free does not optimise for value, and the people who will look worst under it are usually the ones exercising the most discretion. That failure is a close cousin of the training gap that makes so many AI rollouts measure the wrong thing in the first place.

Related Articles

• Only 42 Percent of Middle Managers Feel Ready to Lead
• Are Managers Becoming Obsolete or More Important Than Ever in 2026
• The Career Development Gap Most Workers Blame on Themselves
• Managing AI Agents Is Becoming Its Own Job and Its Own Burnout Risk
• AI Is Hiding a Learning Debt That Could Stall Your Career
• Managers Are Feeding Employee Names Into Public AI Tools
• EY Tells Junior Consultants Remote Work Is Not the Route to Success
• Doctors Say They Should Keep the Savings When AI Makes Them Faster
• The 6 Skills UK Employers Named After a 5-Year Study
• Why Workers Turn Down a Promotion That Costs Them Their Evenings
• Critical Thinking Skills and the Reasoning Habits That Get You Hired
• What Are Humanity Skills and Why They Matter in Your Career

People Also Asked

Q: Does this mean I should produce less work?

A: No, it means volume no longer works as evidence on its own. The argument is that output used to be expensive, so the effort itself signalled that you cared enough to think through what you were putting into the world. Now that a thousand words of coherent prose takes nine seconds, the same volume proves far less. Produce what the job needs, then make the decisions behind it visible.

Q: Which skills are employers actually paying more for?

A: Judgement, creativity and leadership. PwC analysed more than one billion job advertisements and found AI increasing the emphasis on exactly those human skills, with the average wage premium for AI skills rising to 62 percent. Roles where human judgement is emphasised are seeing twice the job growth and 42 percent faster salary growth than roles AI has made easier for non experts.

Q: Is it a bad sign if my employer runs an AI usage leaderboard?

A: It is a sign the metric has not caught up. Disney's internal dashboard ranks product and technology employees on request counts and token use, and one employee invoked Claude about 460,000 times in nine work days. When output is no longer scarce, counting more of it does not identify the people creating the most value. Treat the dashboard as information about how you are currently assessed rather than as a description of what your employer needs.

Future proof your career by getting ahead of what your employer is about to start measuring. Metaintro tracks how AI is changing pay, promotion and performance across real roles, so you can see the shift before your review does. Create your free Metaintro profile and find work that pays for judgement rather than throughput.