AI is redefining how we think about jobs

In 1998, the US Department of Labor set out to answer a deceptively simple question: what do people actually do at work?

Not their job titles. Not their listed responsibilities. The actual tasks, performed by real people, on a typical working day.

The result is a database called O*NET. It took decades to build, it's updated continuously, and it now covers more than 1,000 occupational titles representing over 55,000 jobs across the American economy.

When you look at what it found, something unexpected happens to the way you think about work.

For anyone trying to build an AI-ready organisation, the most interesting thing it reveals is not how different jobs are. It's how similar the work beneath them can be.

Most jobs are made of the same parts

At the title level, jobs look enormously varied. A hospital receptionist and a marketing manager seem to have nothing in common. A financial analyst and a customer success lead look like completely different roles.

But drill down to the task level, what each person actually does across a typical week, and the picture changes.

The same workflows appear again and again across almost every role: updating a record, changing a status, preparing a summary, requesting an approval, scheduling a follow-up, briefing a colleague, triaging a queue.

A patient in a healthcare system and a customer in a billing system are essentially the same entity type. The receptionist updating the patient record and the accounts team updating the customer account are performing structurally identical tasks.

The vocabulary is different. The underlying work isn't.

Research now estimates that 57% of US work hours could be automated with technology that already exists. That number is large because the tasks underneath most knowledge work roles are far more repetitive and cross-functional than the titles suggest.

The title tells you where someone sits. The task tells you what actually happens. Those two things are not the same, and for most of history it didn't matter that we confused them.

Now it does.

Why jobs got bundled in the first place

A typical knowledge work role contains roughly 30 workflows. Some run 20 times a day, others run twice a year. Some require genuine expertise and judgment. Others are essentially data entry dressed up in professional language.

The bundle exists for a practical reason: human beings are linear. One person can only do one thing at a time.

So organisations grouped tasks into jobs, jobs into departments, departments into org charts. It was the only sensible way to coordinate work at scale when human attention was the scarce resource.

That constraint shaped everything: how companies hire, how they structure teams, how they measure performance, and how they think about what a "role" even is. A job wasn't a natural unit of work. It was a coordination solution.

Agentic apps change that constraint in a way no previous technology did.

Tasks can now be triggered, executed and completed in parallel, distributed across humans, systems and agentic apps simultaneously.

Work that once depended on one person moving through a sequence can now be broken apart and handled at the component level.

The bundle made sense when it was the only option. It's not the only option anymore.

What every previous technology missed

It's worth being clear about what makes this different from earlier shifts.

Cloud computing made software accessible anywhere. Mobile made it available any time. SaaS turned expensive enterprise tools into subscriptions. Each of those changed how work got done. None of them changed the structure of work itself.

A spreadsheet is faster than a paper ledger but it still requires someone to enter the data, interpret the results and decide what to do next. Email is faster than post but it still requires someone to read, respond and follow up.

The work got quicker. The underlying workflows stayed the same.

AI is different because it can operate on the task level, not just the tool level. It can draft, classify, summarise, reconcile, recommend and route. It moves through the component parts of a role in a way that cuts across the neat lines of an org chart.

Productivity reports estimate that AI will significantly impact 20% of US jobs by 2030, not by eliminating them wholesale, but by changing the work inside them.

AI replaces tasks, not jobs. But the cumulative effect of enough tasks changing is that the job itself becomes something different.

The design question nobody is asking

When AI changes the task layer, it creates a genuine design decision that didn't exist before.

Which tasks stay human? Which become agentic? Which existed only because information wasn't structured well enough in the first place?

Most organisations aren't asking those questions. They're adding AI tools to existing job structures and measuring how much faster things get done.

That's useful. But it's not transformation. The real opportunity is structural.

It means building shared definitions of what good work looks like, recording how decisions actually get made, and creating orchestration that encodes institutional knowledge rather than leaving it in the heads of whoever has been around longest.

That structure is what a data schema makes possible, and it's what separates organisations that get faster from organisations that get smarter.

In an agentic organisation, the organising unit isn't the job title, it’s the goal.

The job isn't disappearing. It's being redefined.

The O*NET database wasn't built to predict the future of work. It was built to understand the present. But what it reveals is that work has always been more granular, more repetitive and more structurally shared across roles than our job titles ever suggested.

AI doesn't create that reality. It just makes it visible for the first time.

The question isn't whether your jobs will be affected. They will be, at the task level, in ways that are already underway. The question is whether you redesign around that visibility or wait for someone else to do it first.

The org chart was built for a world where humans did everything. That world is changing.

What replaces it starts with a question most organisations have never properly asked: not what are your jobs, but what is the actual work?