If the job keeps changing, learning has to become part of the job

I saw something from The Guardian this morning that looked, at first, like an interesting journalism recruitment story.

The Guardian has launched a Digital Journalist Scheme for people who already have two or three years of experience in journalism, media or content creation. The roles last 18 months and pay £48,000. Candidates only need established experience in one area — text, video, audio, pictures, social or audience — because the point of the scheme is to broaden what they can do.

The part that really caught my attention was this: at least 20% of their working time will be spent on training.

That includes a Level 3 Content Creator apprenticeship, Guardian-designed training, mentorship and practical work across different editorial departments. The Guardian explicitly says the scheme is intended to develop journalists who can work across multiple formats while understanding audience needs, analytics and digital distribution.

I think there is something considerably more important here than a newsroom learning programme. The Guardian is acknowledging, structurally, that the job is changing faster than it is reasonable to expect people to arrive already knowing how to do all of it. A lot of content organisations have the opposite operating model.

The expanding job description

Look at what a good content marketer can now reasonably be expected to understand.

They still need to be able to research something properly, interview somebody, recognise a story, structure an argument, edit a piece and write clearly. But add SEO, audience analytics, social distribution, video, podcasts, customer storytelling, product positioning, commercial context, AI search, AI-assisted research and production, and possibly some understanding of workflow automation.

Increasingly they also need to understand how content travels through a company’s sales, product and customer systems.

Nobody finds that list particularly surprising any more.

What is strange is what often happens next. We put all of it into the job description and go looking for somebody who already has it.

The role evolves. The hiring specification expands.

The organisational investment in developing the person often doesn’t. That increasingly looks like a broken model.

I’ve been writing around this problem for months

I didn’t start with training.

I started with AI changing the economics of production.

In The editorial jobs are not disappearing. They are moving upstream, I looked at a series of jobs asking editorial people to do something broader than produce content: establish narrative, work with evidence, advise leaders, find stories across an organisation, build systems and connect editorial work to commercial strategy.

My conclusion was that as AI makes production cheaper, the valuable human work does not necessarily disappear. It moves. Closer to evidence, decisions, customers and organisational knowledge.

That sounds encouraging if, like me, you’ve already spent years developing those abilities. Then I ran into the much less comfortable question.

How does the next person acquire them?

That became Raising the floor is not building a ladder.

AI can help somebody with less experience produce a much better first draft. I’ve seen that in my own work.

But producing mediocre work and having somebody more experienced explain why it was mediocre was also one of the ways people learned. The boring work was doing two jobs at once. It produced something the organisation needed.

And it produced a more capable worker. If AI removes the first function, we can accidentally remove the second without noticing.

The output improves. The apprenticeship disappears.

We never really designed the old apprenticeship

This is what makes the problem difficult.

Very little of my own development was labelled training.

I learned because editors changed my copy. Because an interview went badly. Because somebody asked why the interesting quote wasn’t in the article. Because I wrote something technically competent that nobody cared about.

I learned because the angle I thought was clever didn’t survive contact with the evidence, and because I watched experienced people make decisions I wasn’t yet capable of making and gradually understood why.

It was inefficient. A lot of it was frustrating. It also accumulated.

Organisations could afford that model partly because they needed the output anyway.

AI changes the bargain.

If a capable system can produce the basic draft, summarise the interview, repurpose the article and give you twenty headlines in seconds, there is a very tempting organisational conclusion: we don’t need somebody spending years learning to do those things.

Perhaps we don’t.

But that doesn’t answer the next question.

What replaces the learning that happened while they were doing them?

This is why the Guardian example interests me

The Guardian’s scheme is not presented as an AI programme.

I actually think that makes it more useful.

It isn’t teaching people to become “AI journalists”. It is recognising that modern journalism spans more surfaces than the traditional role did and building the development of those capabilities into paid work.

You arrive with something. You don’t have to arrive with everything. That’s a very different assumption from the unicorn job description.

The candidate might already be a strong writer. The organisation’s responsibility is then partly to expose them to audio, video, audience development or another adjacent discipline.

Someone from video comes with a different base.

The point isn’t to erase specialism and create identical generalists. The Guardian asks candidates to apply through pathways related to their existing experience and then develop broader capabilities from there.

That feels closer to how capability actually develops.

Depth somewhere first. Expansion from there.

The 20% matters more than the qualification

There is another reason I think the model is interesting. Training has time attached to it. At least 20%.

That sounds almost banal until you compare it with how development operates inside many jobs. Learn AI.

Become better at video. Understand analytics.

Get closer to customers. Learn the product.

Build your strategic capability. But also hit every deadline, attend every meeting and maintain the same production volume.

Development becomes the work somebody is supposed to do after completing the work. Which means it loses.

An organisation can say learning matters endlessly. The calendar usually reveals whether it actually does.

Giving learning a material proportion of paid working time changes the signal.

It says capability development is part of the output of the job.

This connects to a bigger organisational shift

I’ve recently been writing about the idea that AI is changing the organisation around the job. That argument came from looking at companies reorganising management, information flows, job boundaries and individual-contributor roles as the cost of knowledge work changes. Training belongs in the same conversation.

It isn’t enough to give somebody ChatGPT and leave the rest of the organisation untouched. If the technology changes what junior work looks like, the development system has to change.

If experienced individual contributors become more valuable, career structures have to change. If jobs span more disciplines, learning has to change.

If AI can raise the quality of an inexperienced person’s output before their judgement has caught up, assessment has to change. The technology changes the job.

Eventually the organisation around the job has to respond. The Guardian scheme looks like one version of that response.

Content marketing has this problem badly

I think B2B content teams should pay particular attention. Content marketing has spent years accumulating responsibilities. First writers needed to understand digital.

Then search. Then social.

Then demand generation. Then multimedia.

Then product marketing. Then data.

Now AI. The usual response has been to make the role wider.

I increasingly think the more interesting response is to make the development system better. Instead of asking:

Can we find a writer who understands our industry, SEO, video, social, analytics, AI, product, customers and commercial strategy?

Ask:

What does this person need to be excellent at when they arrive, and what are we prepared to help them become excellent at next?

Those are radically different hiring questions. One searches for finished capability. The other assumes capability compounds.

AI makes this more urgent, not less

I wrote in AI didn’t replace editorial judgement. It made it more valuable that AI now touches most of my own workflow. It can help me interrogate research, compare transcripts, test arguments, structure material and produce drafts.

That has not reduced the amount of judgement involved. It has increased the number of things I can exercise judgement over.

But there is an enormous privilege hidden inside that argument. I’ve already had the career.

I already know what a weak interview feels like. I already know that a plausible statistic can still be the wrong statistic.

I already know that the first angle is often not the interesting one. I already have enough scar tissue to distrust something that looks finished too quickly.

Someone entering the profession now gets the machine before accumulating all of that. Telling them to “use AI well” isn’t a training strategy.

I tried to answer this once already

That was the problem behind The training gap: AI has to be in the curriculum. My argument there was that AI training shouldn’t be a prompting course sitting beside the real craft. It has to happen inside the craft.

Give somebody machine-generated work and teach them to diagnose what is wrong with it. Make them defend the evidence.

Make them explain why they changed the argument. Make them identify what is missing.

Have somebody experienced interrogate those decisions. The expensive part is that last sentence.

Judgement still needs contact with better judgement. You can automate the first draft.

You cannot simply automate apprenticeship and assume the same professional comes out the other side. The Guardian scheme interests me because it appears to accept that development needs both real work and deliberate instruction.

There are reasons not to overstate this

One scheme at one newspaper does not prove a new employment model.

The positions are fixed term. We don’t yet know what happens to participants afterwards. Setting aside 20% of someone’s time does not guarantee that operational pressure will never invade it. A formal qualification can be excellent or bureaucratic depending on how it is delivered.

And The Guardian is a large newsroom with very particular capabilities. A five-person B2B marketing team cannot replicate its internal rotations, specialist desks and training infrastructure.

Those limitations matter.

What can travel is the principle.

Stop treating development as something employees are responsible for fitting around production.

If organisations require people to keep acquiring new capabilities because the work keeps changing, then some of the cost of acquiring those capabilities belongs to the organisation.

Perhaps this is the new side of the bargain

There has been a lot of discussion about what AI should allow companies to expect from employees. More output. Broader roles.

Faster execution. Smaller teams.

More autonomy. Those expectations may all turn out to be reasonable in some situations.

But there should be another side to that bargain. If one person’s potential surface area expands, organisations should invest in helping that person grow into it.

Protected learning time. Real projects rather than abstract courses.

Access to experienced practitioners. Feedback on judgement rather than only output.

Permission to develop adjacent skills without already being an expert in them. A pathway from one capability into several rather than a requirement to possess everything on day one.

That seems particularly important for content work because the job is not becoming simpler. Production is becoming easier.

Those are not the same thing. The Guardian appears to understand that distinction.

A good content organisation shouldn’t just ask how AI changes what one person can produce. It should ask something harder.

What kind of organisation would help that person become capable of doing the work that matters next?

That might be the more important redesign.