AI lets you work in the order the problem demands

The order isn’t neutral. It’s usually inherited from whichever discipline happens to go first — and AI is one of the few things cheap enough to let a different order win.

One of the less obvious things AI is changing is the order in which work gets done.

Content production inside organisations often follows a fairly fixed sequence. A brief is agreed. A format is chosen. A page or a campaign is designed. Writers are given sections to fill. Copy is squeezed into the structure that already exists.

There are good reasons for working this way. Different disciplines have different skills, design systems exist for a reason, and someone has to own the build.

But a project I’m developing at the moment has worked almost in reverse. I can’t share the details until it’s published and live, but the shape of the process is worth setting out now, because I think it generalises well beyond this one piece of work.

Starting from the evidence, not the format

The project is based on a series of customer interviews, so it started with a lot of raw material rather than a brief: transcripts, customer evidence, different experiences, quotes, repeated themes, contradictions, potential stories.

Traditionally, getting from that volume of source material to a coherent editorial argument can take a long time. You read every transcript. You make notes, compare interviews and start grouping themes. You go back looking for supporting quotes, then change your mind about the structure. You realise that something interesting mentioned in interview four connects with something somebody said in interview one, and the grouping shifts again.

That work is valuable, but it is also cognitively expensive — which is exactly why it so often gets skipped or shortened under deadline pressure.

Compressing the distance from information to judgement

AI changes the economics of that stage quite dramatically. I can give it large amounts of source material and use it to interrogate what’s actually there: which themes recur, where customers disagree, which statements have real evidence behind them, which examples support a particular lesson, where two interviews are describing essentially the same problem in different language, and what potentially interesting material I’m overlooking.

That doesn’t mean asking AI to tell me what the story is. That distinction matters. The editorial judgement still sits with me — deciding what’s significant, what’s credible, what the argument should be, what to leave out and what the audience might actually care about. But AI can dramatically compress the journey from information to something I can exercise judgement over. For this project, I used that process to identify the strongest themes, connect them to supporting evidence and develop the editorial hierarchy.

The content started designing the page

Only once that hierarchy existed did the shape of the experience start to become obvious. It needed certain sections. Certain pieces of evidence needed more prominence than others. Video belonged in particular places. Some ideas required more space than others. The brand needed to be present, but as an evidenced enabler rather than something pasted over the top of the story.

The content, in other words, effectively started designing the page. AI helped there too — not because it suddenly made me a web designer, because it didn’t. What it gave me was enough access to an adjacent discipline to explore the problem properly. I could test structures, explore hierarchy, modules and page rhythm, and move material around before anyone had committed significant time to building it. I could turn an editorial argument into something close enough to an experience that a web specialist could then take it and work out how to implement it properly inside the real system.

That distinction matters.

Expertise doesn’t disappear — the boundaries get more permeable

There is a very easy AI story to tell here: that AI means everyone can now be a writer, designer, developer, researcher and strategist all at once. I don’t think that’s what’s actually happening.

Expertise still matters. AI can find patterns in transcripts without understanding which ones matter politically, commercially or editorially. A prototype can look convincing while containing poor UX decisions, accessibility problems or things that simply don’t work inside a real design system. The professional disciplines haven’t disappeared.

What AI may be doing instead is making the boundaries between them more permeable:

• A writer can interrogate a body of research more like an analyst.
• A researcher can explore how evidence might become a story.
• A writer can temporarily think more like a content designer.
• A strategist can test an execution rather than stopping at a slide.

You don’t necessarily become the person who owns the adjacent discipline. You gain enough range to move much further through the problem before specialist intervention becomes necessary — and that changes something more fundamental. You don’t always have to work in the order the organisation was designed around. You can work in the order the problem demands.

Two sequences, not one universal rule

For this project, the sequence has been something closer to: source material → AI-assisted synthesis → editorial judgement → evidence → narrative → structure → copy → prototype → professional build.

Rather than the more usual: format → design → content slots → copy.

Neither sequence is universally correct. Sometimes design should lead. Sometimes a standard template is exactly what a piece of work needs. Sometimes the smartest editorial decision is to fit neatly into an existing system rather than invent anything new. But AI gives individuals much more flexibility to test alternative sequences cheaply, and I think that matters for editorial work in particular.

Format-first is how mediocre content gets made

A lot of mediocre business content starts with the asset: “We need an article.” “We need a landing page.” “We need a video.” “We need six social posts.” The format gets chosen before anyone has really decided what’s worth saying.

I’m increasingly interested in reversing that — starting with the information, interrogating it, finding the evidence, working out what you actually know, identifying what’s missing and making the editorial judgement, then determining what the thing should become. AI makes that approach much more practical. It reduces the cost of processing large amounts of information and gives you enough capability to explore several possible forms before specialist production starts.

Specialists still finish the job

That doesn’t remove the specialists — if anything, it can make their involvement more valuable. Instead of asking a designer or developer to solve an unclear editorial problem, you can give them something where much more of the underlying thinking has already been worked through.

The interesting impact of AI, then, may not be that it collapses disciplines. It may be that it collapses some of the walls between them, and dramatically shortens the distance between raw information and a decision about what to do with it. Once that happens, the workflow doesn’t have to begin with whoever normally goes first. It can begin with the part of the problem that actually needs solving.