The framework disappeared into the work

I came back to work at the beginning of September with the usual problem after a long break: too many threads, too much context and no reliable way to know which of the things that looked urgent before the holiday still mattered.

A few years ago I probably would have started with the backlog.

This time I did something different. I reconstructed the state of the work first. What had moved? What had died? What was waiting on somebody else? What new evidence existed? Which apparent content requests were actually research problems, coordination problems or decisions that had not yet been made?

I did not sit down and decide to “use Editorial Intelligence” at Sage. I barely thought about it in those terms. But looking back over the first two weeks, a lot of the ideas I have been developing through Editorial Intelligence had quietly become the way I was doing the job.

That may be a better test of a framework than whether you can explain it on a diagram. Eventually the framework should disappear into the work.

Catching up became an information problem

The first change was small but important. I stopped treating return-to-work as a task-management problem and started treating it as a reconstruction problem.

Before producing anything, I wanted to know what had changed while I was away, what had already been progressed, what was still required, which decisions were unresolved and where my involvement would add something rather than duplicate somebody else’s work.

That is closely related to the idea behind making work recoverable. The point is not to preserve every task forever. It is to preserve enough context that you can resume the work intelligently.

The interesting bit is that I was no longer consciously applying that principle. It had just become how I approached the problem. The broader Editorial Intelligence in practice page describes the same pattern: frameworks emerging from repeated practical problems rather than being imposed on the work from outside.

GEO stopped meaning “make another article”

The clearest example has been GEO. The old content reflex is easy to recognise:

a question appears → commission an answer

What I have found myself doing instead is closer to:

a question appears → inspect the current answer environment → diagnose what is actually missing → only then decide what to do

Sometimes the gap is an answer. Sometimes it is evidence. Sometimes the answer already exists but is buried in the wrong page structure. Sometimes what is missing is practitioner proof, customer experience, product explanation or a clearer direct response.

Using Profound made that distinction much more concrete. A broad topic can look healthy while a very specific question remains weak. But even a weak question does not automatically justify another article. The web may already explain the definition perfectly well. What it lacks might be the practitioner layer: what people misunderstand, where the work actually appears, what changes operationally, and how somebody doing the job would explain it in plain English.

That is the point where GEO became interesting to me as a reporting discipline, rather than simply an optimisation discipline. It is also the distinction behind the AI Search Visibility framework and the more practical AI Visibility Audit: visibility problems need to be separated into discoverability, answer quality, evidence, authority and genuine knowledge gaps before deciding what to make.

GEO became useful when it stopped being a distribution problem and started becoming a reporting brief.

An event became an evidence engine

That changed how I started thinking about Accountex Manchester. Instead of going with a generic instruction to “get some content”, I now have a set of known evidence gaps I can investigate while I am there, which gives the event a more useful job than simply generating assets.

The chain is becoming clearer:

GEO identifies a weak question → editorial defines the missing evidence → the event gives access to practitioners → written, video, social and web experiences can use that evidence differently → the same question can be measured again afterwards.

I wrote more specifically about what I am looking for at Accountex Manchester, but the wider point is about what an event is for. An event is obviously a production opportunity: you can film, interview, post, recap and distribute. It is also a temporary research environment where customers, practitioners, experts, internal specialists and live industry discussion are concentrated in the same place.

If you already know what your organisation does not know, that room becomes much more valuable. That is the Evidence Engine idea showing up in practice without needing a workshop about the Evidence Engine. It is also close to the logic of research-led thought leadership: start with the question and the missing evidence, then decide what the evidence should become.

The channels stopped looking like separate jobs

Another thing I have noticed is that social, editorial, customer stories, video and GEO are starting to look less like separate content jobs.

I work across social content and customer stories, while a lot of my own background is rooted in written editorial. That means I can see the same underlying idea appear in very different forms.

A research narrative might become simpler language on social. A practitioner interview might expose the evidence behind it. A longer article can establish the fuller answer. A short video can provide the human proof. GEO then asks whether the organisation has made those answers clear and authoritative enough to surface when somebody asks an AI system for help.

The thing travelling between those channels is not the asset. It is the evidence and narrative underneath the asset. That is the same problem addressed by the Content Activation Framework: how one evidence base can create multiple useful experiences without turning every channel into a duplicate of every other channel.

That has changed the way I think about consistency too. Different channels do not need identical wording. They need enough conceptual coherence that you can still recognise the same underlying territory.

The asset changes. The knowledge should survive.

Publishing stopped looking like the end of a customer story

The same pattern has appeared in customer work. I have been experimenting with what happens if a customer story does not simply disappear into the archive after publication.

What evidence did it contain? What was the story spine? Which proof points mattered? Which quotes were genuinely useful? What did the editorial process learn? What gaps remained? What should the next interviewer know because this story happened?

That does not mean treating customer stories as a database for its own sake. The point is to stop useful customer intelligence evaporating when the asset is finished. The Customer Signal Framework and Case Study Intelligence Framework are both attempts to make that learning reusable rather than letting every customer conversation start from zero.

There are important boundaries. Approved customer evidence is not the same thing as internal editorial learning, and one customer’s claims do not become evidence for another customer. But the larger shift still matters:

Publishing stopped looking like the end of the customer story. It became one point in a longer memory loop.

AI widened the analysis. Judgement narrowed the action

A messier campaign project made another principle more obvious. There was a lot of material — audit work, possible content gaps, different campaign phases, different owners, evidence opportunities and plenty of potential things that could be made — but the value was not in generating more ideas.

It was in absorbing the complexity, spotting duplication, working out what could be reused or merged, understanding what actually mattered now, and turning a large internal analysis into a much smaller recommendation somebody could act on.

I keep coming back to this pattern:

Use AI to expand the internal analysis. Use editorial judgement to compress the external recommendation.

That feels much closer to the useful role of AI in knowledge work than simply “AI made me faster”. The gain is not output volume; it is reducing the amount of unresolved complexity you hand to somebody else. The service version of that logic is often a Narrative Audit: understand the research, customer insight, expertise and content already present before deciding what needs to be created next.

Protect the idea, not every artefact

Even a fairly ordinary web-production decision reinforced the same thing.

An editorial concept moved into a real CMS. Some of the presentation assumptions changed. The hero treatment did not need to survive exactly as imagined. Video did not need to appear everywhere. The implementation had constraints the original mock-up did not.

What mattered was preserving the underlying editorial idea rather than defending every design choice around it. I wrote down the lesson as:

separate the editorial requirement from the presentation preference. Protect the idea, not every mock-up choice.

That is another version of the same pattern: preserve the durable reasoning while allowing the implementation to change.

The unexpected part

I had assumed that taking Editorial Intelligence into a large organisation would eventually mean persuading people to use some of its frameworks. So far, the opposite has been more interesting: the useful principles have travelled best when I have not needed to name them at all.

They show up as a better question. A smaller brief. A decision not to make something. A piece of evidence captured before it disappears. A clearer handoff. A recognition that a visibility problem is actually a knowledge problem. A refusal to create another process when the existing one can simply make a better decision.

That connects to what happened with the EI Router. The router got worse when it started manufacturing work for itself. The correction was simple: routing should make complicated work simpler. It should not make simple work complicated.

The same rule seems to apply to the whole system. Editorial Intelligence is less useful to me when it becomes ceremony; it is more useful when it changes the judgement almost invisibly. That is also why I am interested in the editorial judgement system audit as a service: not to automate editorial judgement, but to make recurring useful judgement easier to retain and reuse.

I am not claiming that this proves the framework. It is too early for that, and some of these experiments may still fail.

Accountex will be one useful test. Does going in with known evidence gaps actually change the conversations? Do practitioner answers reveal something the current answer environment is missing? Does the same evidence naturally find uses across written content, video, social and web? Do any of the assumptions collapse once they meet the real room?

I can write that part afterwards. For now, the thing I find most interesting is that I have stopped noticing some of the method while I am using it.

The method becomes judgement. The judgement becomes habit. And eventually the framework disappears into the work.

A framework probably becomes useful at the point you no longer need to announce that you are using it.

Where this sits in Editorial Intelligence

This diary entry is the practical overlap between the Editorial Intelligence Framework, Evidence Engine, Customer Signal Framework, Case Study Intelligence Framework, Content Activation Framework and AI Search Visibility framework.

For organisations facing the same kinds of problems, the closest service routes are the Narrative Audit, research-led thought leadership, AI Visibility Audit and the wider Editorial Intelligence consulting offer. The point of those services is not to install more process. It is to understand what the organisation already knows, identify what it genuinely needs to learn, and turn that into evidence, narrative and editorial capability that can compound.