AI didn't replace editorial judgement. It made it more valuable.

Everyone seems to ask the same question: how much of my work does AI do? My answer usually surprises people — probably around 80% — but that’s also the wrong number. AI doesn’t do 80% of my work. It touches around 80% of my workflow, and the difference matters.

Almost every stage of the editorial process I follow now has an AI-assisted step: research, transcripts, customer interviews, presentation drafts, theme extraction, comparing evidence, structuring ideas, testing alternative narratives. The work hasn’t disappeared. It has changed, and the more AI becomes involved in my workflow, the more valuable editorial judgement has become.

The work AI actually changed

A few years ago, producing a substantial piece of thought leadership looked something like this: read the research report, read every interview transcript, highlight the key quotes, compare notes, build a presentation, write the article, create social posts, repeat. Much of that effort wasn’t creative. It was the mechanics of knowledge work — reading, organising, comparing, restructuring and searching for information that already existed somewhere in the organisation. Those activities mattered; they simply weren’t where the greatest value was being created.

Today, much of that mechanical work happens dramatically faster. Not because AI thinks for me, but because it removes much of the friction between gathering evidence and making decisions.

My workflow today

During a recent two-week sprint, I worked across research, customer stories, event discussions, presentations, website content and editorial planning. AI appeared throughout almost every project, helping structure transcripts, compare recurring themes, organise research, draft presentation outlines, identify repeated patterns and suggest alternative ways of framing the same information.

What it didn’t do was decide what actually mattered. That remained entirely human, and if anything AI increased the number of editorial decisions I had to make. Instead of asking what should I write, I found myself asking which of these competing interpretations is genuinely supported by the evidence. That’s a very different job.

One week that couldn’t have happened three years ago

Looking back, what strikes me most isn’t that I used AI a lot. It’s the amount of evidence I was able to work with. Within the same sprint, I was moving between research reports, customer interviews, industry roundtables, paid campaign performance, presentation design and website content. Individually, none of those projects appeared connected, and historically they would probably have stayed separate — research informing one article, an event generating another, customer stories becoming case studies, performance data sitting inside a dashboard.

Instead, AI made it practical to compare all of them together. As the evidence accumulated, the same operational pressures kept appearing in different places: research pointed in one direction, customer interviews reinforced it, industry discussions expanded it, performance data suggested audiences responded to it. AI didn’t discover a hidden truth. It dramatically reduced the effort required to compare evidence that would previously have lived in different folders, presentations and meeting notes. The editorial work came afterwards, recognising that those separate signals were actually describing the same story.

The narrative wasn’t sitting in one document

One project involved bringing together multiple evidence sources around the changing role of accountants — research findings, customer interviews, industry event discussions, paid media performance, internal conversations. Individually, each source told part of the story; none contained the complete narrative. Only after comparing them together did something become obvious: the same operational pressures kept surfacing regardless of where the evidence came from.

That insight wasn’t generated by AI. AI accelerated the comparison; editorial judgement recognised that a genuine pattern existed. The resulting narrative wasn’t simply a summary of the research — it was an interpretation supported by multiple independent sources, and that’s a very different form of work.

A few days later, I was analysing discussions from an industry roundtable about embedded services and AI. Again, AI helped process transcripts and organise themes, but something more interesting happened: the conversations weren’t isolated. They echoed many of the same operational challenges I’d already seen elsewhere. What initially looked like a separate project became the next chapter of the same narrative — not because AI joined the dots, but because editorial judgement recognised that different evidence sources were describing the same underlying change from different perspectives. That’s becoming an increasingly common part of my work: not creating information, but connecting it.

AI creates possibilities. Editorial judgement creates direction.

One thing I’ve noticed is that AI rarely struggles to generate ideas. Ask it for headlines and you’ll get twenty; ask it for article structures and you’ll get ten; ask it for strategic recommendations and you’ll probably get five sensible answers. The bottleneck isn’t generating possibilities any more. It’s deciding which possibility deserves to survive. That’s the shift I’ve experienced: AI accelerates synthesis, editorial judgement determines significance.

Multiple models rarely agree

Different AI models regularly produce different interpretations of the same material — one may simplify too aggressively, another may overcomplicate the argument, a third may confidently recommend an entirely different direction. None of those models knows which interpretation is correct; they generate possibilities, not truth. That’s why I rarely accept the first convincing answer. I compare different outputs against the available evidence until a consistent picture begins to emerge. Editorial judgement isn’t choosing your favourite answer. It’s deciding which answer is best supported by the evidence.

AI removed friction, not responsibility

People often assume AI has replaced writing. I don’t think that’s what has happened. What AI has really replaced is much of the administrative work surrounding knowledge creation — searching, sorting, formatting, comparing, drafting first versions. Before AI, connecting ten sources of evidence might have taken days; today it can take hours.

That doesn’t reduce the need for editorial judgement. It raises the standard for it, because once processing information becomes easier, the quality of the decisions becomes the competitive advantage.

This is why I started thinking differently

The more I worked this way, the more I realised I wasn’t simply using AI more effectively. I was beginning to recognise the outline of something larger: a different way of working. I think Editorial Intelligence is becoming a discipline in its own right — not because AI created it, but because AI has fundamentally changed where human expertise creates value. Every discipline emerges when technology changes the nature of work, and AI has dramatically reduced the cost of processing information without reducing the need to interpret it. If anything, it has made interpretation more valuable.

At its simplest, Editorial Intelligence is the discipline of turning research, customer insight, executive expertise, market signals and AI into better editorial decisions. Its purpose isn’t to produce more content. It’s to improve the quality of organisational thinking before content is ever created. AI makes that discipline possible at a scale that would have been impractical only a few years ago, letting us compare more evidence, test more hypotheses and surface more patterns than manual workflows ever could.

But AI doesn’t replace judgement. It increases the importance of it, because AI can process evidence, and only people can decide what that evidence means.

The future isn’t better prompting

There’s a great deal of discussion about prompts, agents and automation, and those conversations matter, but I don’t think they’ll become the defining capability. The defining capability will be judgement — not because AI fails, but because AI succeeds. As models become better at generating information, information itself becomes less scarce, and editorial decisions become more scarce. AI is changing who can create content; Editorial Intelligence is about deciding what deserves to be created. Those aren’t the same thing.

The organisations that thrive over the next decade won’t necessarily produce the most information. They’ll make the best decisions about which information becomes action. That’s why I believe Editorial Intelligence is emerging as a discipline — not because AI replaces editorial thinking, but because AI makes editorial thinking the point of greatest leverage. The future won’t belong to the people who generate the most possibilities. It will belong to the people who consistently recognise which possibilities are actually true.