What even is a job anymore?

Hello again my friend,

There’s this guy, good friend actually, and he works in accounting at one of those enormous firms with thousands of employees. It’s one of those cases where someone thinks they’re smart until they realize they’re on the verge of losing their job.

He’s senior enough that around half of his day is spent in meetings. The rest is reports, document work, and the technical stuff you would expect from someone who has spent years working his way up at a major accounting firm.

Recently, he described his new routine to me:

He was proud of it too! '“ahaha so easy,” “nobody even notices, LOL,” “I get my entire day back hoho.” He wouldn’t stfu it was so annoying. But I get it. It feels good.

Then he asked me what I would do with all that extra time.

I told him he had two good options. He could pick it up at work and become considerably more valuable (make more there), or he could finally start the business he keeps talking about and go all-in while keeping the day job.

He was surprised by the first part. Why would he volunteer to do more work when everything expected of him was already getting done so quickly and easily?

Fair. If his company is happy, and he is meeting every expectation, I don’t think he should just sit there and jerk his mouse around a few times so Slack shows he’s ‘online,’ while doing nothing. Getting time back is awesome in some cases.

But, enjoying the extra time, while betting they never notice is dangerous. He’s very obviously taking advantage. I told him very clearly that his name will be added to the layoff list the moment someone important finds out.

The question is how long that lasts.

If a director can automate most of his reports, a product manager can design, a salesperson can code, and almost anyone can produce a clear, coherent document, what exactly is a job anymore?

Let’s lock in.

For most of our careers, a job title came with a fairly clear set of boundaries.

Designers designed. Engineers coded. Salespeople sold. Product managers wrote requirements and kept everyone moving. If something crossed into another department, it was handed over to the person whose title matched the work.

That system made sense when the ability to do each kind of work was difficult to acquire. If a product manager wanted to build a feature themselves, they needed years of engineering experience. If an engineer wanted to design something, they needed to learn the software, the principles, and probably how to make the whole thing not look like shit.

Now, you can describe what you want in plain English and get a reasonable version in an afternoon.

It may not be brilliant. It may not be ready to ship. But it is often enough to keep the work moving without waiting two weeks for someone in another department to have time.

I’ve already written about the player-coach model, where managers still lead but also do the work. I’ve written about roles moving sideways, where product managers learn to code, designers develop product judgment, and engineers get closer to customers.

It’s more obvious to me now that they’re expressions of the same phenomena.

A title is becoming less like a boundary and more like a position on a basketball team.

A point guard will probably dribble and pass more than everyone else. A center has a very different job around the basket. Some players are on the floor because they are brilliant shooters.

But everyone still plays defence. Everyone passes. Everyone has to shoot a free throw when they’re fouled.

Being a great shooter does not mean you take every shot. It means that when the shot matters, the team tries to get the ball into your hands.

Work is moving in the same direction.

The brilliant designer should not have to create every small internal page or one-off sales asset. Other people can take those shots now. But when the company is redesigning its most important product, changing foundational customer behaviour, or setting the standard future AI agents who design, you probably want the ball in that designer’s hands.

Specialists still matter.

The difference is that being a specialist no longer gives you an excuse to stand still while everyone else plays defence.

This brings up an uncomfortable question for people who have spent years getting good at something.

I don’t think the answer is to deny what is happening. Anyone can “code” now, at least in the loosest definition of the word. I’ve built real software, and I am not going to pretend I suddenly became as good as the engineers I work with. I didn’t.

Technically though, I am producing code.

That distinction matters.

A salesperson can ask Claude to build an internal tool. A product manager can build a working prototype. A founder can launch an entire app without hiring someone first. But producing something that looks like code is not the same as understanding how it fits into a large system, knowing what could break, keeping it secure, or being the person everyone calls when something goes wrong.

The same is true in design. I can make something that looks pretty good. I can share screenshots, explain the vibe, reject a few versions, and eventually get close to what I want. But I don’t have the eye of someone who has spent ten years learning typography, interaction, hierarchy, and how small design decisions change what people do.

AI gives me access to their tools and some version of their output.

It does not give me all of their judgment.

This is where specialists become more valuable in a different way. They set the standard. They review what everyone else produces. They coach the people around them. They handle the strange cases where the normal answer does not work. Most importantly, they understand the consequences of getting it wrong.

Anyone can take the shot. Should they take that shot?

That is also how I would think about raises and promotions now.

The engineer’s value cannot only be the volume of code they personally type. It is the systems and outcomes they can be trusted to own, the difficult decisions they make, the problems they catch before anyone else sees them, and how much better the rest of the team becomes because they are there.

The designer’s value cannot only be how many screens they produce. It is whether the product feels coherent, whether customers understand it, whether the experience helps the business, and whether everyone else learns to make better decisions from working with them.

AI makes production easier. It also makes it easier to produce an incredible amount of bad work that looks acceptable at first glance.

Someone still needs to know the difference.

There are parts of business I think almost everyone should experience now.

Everyone on a software team should answer customer support tickets sometimes. Everyone should be able to explain to a prospect why their company is a better choice than its competitors. Everyone should have enough access to the numbers to understand how the business makes money. And yes, everyone should probably try building an internal tool when they see something repetitive that could be easier.

That does not mean a software engineer should spend half the week doing sales calls, or that a salesperson should be pushing unreviewed code into production. We do not need to turn every company into a group project where everyone does everything badly.

But there is a difference between specializing and hiding.

An engineer who can talk to customers, understand the finances on a spreadsheet, and recognize when a product decision makes no sense is incredibly valuable. So is the designer who can explain the product to a prospect, or the salesperson who understands the product deeply enough to give useful feedback instead of forwarding every request they hear.

We have all worked with exceptionally talented people we would never put in front of a customer. We used to work around that.

They could be brilliant at one specific thing, pass everything else to other departments, and spend most of their career inside that protection.

Unless that person is truly one of the best in the world at a specialization the market desperately needs, I think that road gets much harder after 2026.

The new well-rounded business person still has a specialty. But they also understand customers, money, priorities, communication, and how their decisions affect the rest of the company.

They can step back far enough to see the business while staying close enough to the work to be useful.

That combination is difficult to fake with a prompt.

This brings me back to my friend.

The risk is not that he learned how to complete his work in less time. That part is good. Every company says it wants efficiency, until the efficiency gives someone a free afternoon, (but that’s another conversation).

The risk is that most of the work supporting his senior title can now be completed by AI, and he is treating the difference as permanent free time instead of a temporary advantage.

Markets tend to notice that kind of gap.

Maybe the firm eventually combines his responsibilities with someone else’s. Maybe a younger person who is just as comfortable in the meetings uses AI to complete the documents and takes on two additional areas of the business. Maybe the firm realizes it does not need to keep paying director-level compensation for reports that take an afternoon to produce.

None of that happens immediately. Large companies are slow. They have old processes, complicated politics, and enough money to carry inefficient roles for years.

A smaller company trying to win will not have the same patience.

If one employee uses AI to finish their old job by lunch and goes home, while another uses the same tools to finish the job, improve the system, speak to customers, and take responsibility for something new, it is not difficult to imagine who becomes harder to replace.

That does not mean we need to donate every hour AI saves back to our employers. I would not tell anyone that.

Use some of it to rest. Use some to spend time with your family. Use it to build your own business, learn something, or create another source of income that does not depend on your company deciding to keep you.

But use it deliberately.

Because once a company understands that the old job takes two hours, it is not going to keep paying for the other six forever out of kindness.

There is data starting to show what replaces the narrow task-based job.

AI Jobs Barometer | PwC analyzed more than a billion job postings across six continents. Their researchers found the job market splitting in two directions.

Some jobs are being “democratized.” AI is making the work easier for people without years of specialized experience.

Other jobs are being “professionalized.” AI handles more of the routine work, but the person is expected to bring more judgment, leadership, creativity, and expertise to everything that remains.

The professionalized jobs are growing twice as fast, with 42% faster wage growth since 2021. New tasks added to AI-exposed roles are 2.5 times more likely to need skills like empathy, judgment, and creativity.

One finding really stood out to me: junior roles exposed to AI are seven times more likely to ask for skills we traditionally expected from senior employees, including leadership and strategic thinking.

Think about how strange that is.

The entry-level work is disappearing first, but companies still want entry-level people. They just want them to show up with the judgment we used to let them spend five or ten years developing.

That expectation may be unfair. I’m not sure how someone is supposed to learn senior judgment if nobody will hire them to make junior mistakes. Companies will have to figure that out too.

But it shows where the value is moving.

AI at Work Report 2025: How GenAI is Rewiring the DNA of Jobs - Indeed Hiring Lab and found that 46% of the skills in a typical U.S. job posting are ready for what it calls “hybrid transformation.” AI can perform a large portion of the routine work, but humans are still needed to handle exceptions, interpret unclear situations, validate the output, and accept the legal or ethical responsibility.

Only 0.7% of the skills they studied were considered very likely to be fully replaced.

So the job is not disappearing all at once.

It is being hollowed out and rebuilt around the parts where a person still needs to think, decide, care, and take responsibility.

To understand where this goes, we probably need to look beyond what most people are doing with AI today.

The winning companies will not be groups of people sitting around sending random messages to Claude all day. The setup will be more organized than that.

You might have one agent doing competitor and market research. Another takes that research and drafts the plan. A third reviews it, finds weak points, and recommends changes. Once you approve the plan, another agent creates the first design. Claude builds it. Codex reviews the code. Tests run. The results come back to you.

Somewhere in that process, you might tag the brilliant designer because this part matters enough to deserve their eye. Someone else might tag you because they built something for your customers and you understand those customers better than they do.

You will be working or managing ten agents, fifty, or one hundred even. That is not the interesting part to me. The number will depend on the company, the work, and what all this costs.

The bigger change is that we will be responsible for far more work than we personally completed.

Every agent will need context. Someone has to decide what good looks like. Someone has to review the plan, notice when the research missed something, and stop a bad idea from becoming fifty completed tasks.

And agents do not magically learn because you told one of them it was wrong once.

If the feedback disappears inside a conversation, it can make the same mistake next week. So you need agents that learn (you’ll probably have them). Beyond that, the company has to capture what happened, improve the instructions, fix the underlying data, add a test, or change the process.

The foundations matter. The feedback matters. Prioritization matters even more because agent work is not free, and a company can burn an incredible amount of money producing things nobody needed.

Most of all, someone has to own the result.

You may not have personally researched, designed, coded, and reviewed every piece.

You are still responsible for what ships.

At the beginning of this, I described a world where almost anyone can do almost anything to some degree.

That is already happening.

A product manager can design. A salesperson can code. A receptionist can build a 3D model. Someone who struggled with writing can produce a clear document. A small team can operate with capabilities that used to require an entire company.

But “I can make one” and “you should trust me with this” are very different statements.

A job is becoming less about the tasks that belong to you and more about the outcome you are responsible for. The "worker bee” era is slowly fading away, unless you’re in the top 10%.

A title tells the team where you are strongest. It tells them where your judgment is deepest, where you should coach, and when they should get the ball into your hands.

It does not give permission to ignore everything else happening on the court.

The people who do well in this version of work will probably have a few things in common:

Jobs are not guaranteed. They probably never were, but it felt easier to believe they were when expertise took years to access and every department protected its own work.

Now the boundaries are fading.

I don’t think the answer is to become average at everything. The best teams will still need brilliant engineers, designers, salespeople, accountants, and operators. They will just need those people to contribute beyond the small box that used to sit underneath their title.

On a basketball team, your position still matters.

But the job is to help the team win.

Thanks as always for reading.

Darwin

Sign up for a free trial of Attio, the agentic CRM.

Ask Attio to build a daily workflow that surfaces the deals that need your attention today, like anything with a stage change, a recent reply, or a new signal in the last 24 hours.

Review your pipeline in Claude, synced live from Attio via MCP.

That's it.

Try Attio Now