Most organisations aren't confused. They're comfortable.
Confusion creates urgency. Comfort creates meetings. Meetings create the feeling that something is happening. And the feeling that something is happening is, for most businesses, a perfectly adequate substitute for something actually happening.
This is the fog. Not ignorance. Not incompetence. Not even resistance to change. It's the normalised state of operating without clear intent. Coordination mistaken for progress. Activity mistaken for strategy. Roles and processes that exist not because they produce something specific, but because they have always existed.
The fog is comfortable because everyone is in it together. Every organisation can name what it produces: products, services, programmes, platforms. But could yours tell you, precisely, what outcome it is committed to delivering and how it would know if it had failed? Not what it hopes for. Not the vision statement. The falsifiable commitment.
Something like:
"By June 30, we will cut onboarding time from 14 days to 48 hours without increasing support tickets per customer. If we miss either number, we failed."
If defining that feels uncomfortable, that's the fog doing its job. It protects you from measurement. It protects everyone from it.
I spent years inside it.
What the Fog Looks Like from the Inside
You wouldn't describe it as vagueness if you were living in it. It doesn't feel vague. It feels like work. Hard work, in many cases. Long hours. Genuine effort. People who care deeply about their craft.
The question is whether the systems they operate inside still connect to the people they were originally built to serve.
That disconnection is the signature of the fog. Not laziness. Misalignment so normalised that nobody notices it anymore.
Industries that started to serve people but turned inward. Organisations that measure activity because outcomes are harder to define. Teams that respond to "this isn't working" by adding more process.
Bolt something new onto something old and call it progress.
At first, the seams show. Then you stop noticing them. A tolerance builds until the tolerance itself becomes the fog.
I have watched organisations launch transformation programmes that transformed nothing except the job titles of the people running them. None of these people were stupid. None of them were cynical. They were in the fog, doing their best inside a system that rewarded activity over clarity. And the system felt so permanent that questioning it seemed naive.
There's a particular feeling that comes with this. Not frustration exactly. More like a low-grade suspicion that things should be landing but aren't. That the effort going in is not proportional to the impact coming out. That somewhere between the strategy and the execution, the intent gets lost. Not dramatically. Not in a crisis. Just gradually, persistently, in the ordinary course of doing business.
You've probably felt it. The meeting that could have been an email. The email that could have been a decision. The decision that could have been made six months ago but instead got routed through three more committees and emerged as a compromise nobody believed in.
Here is a question worth sitting with: if you measured your organisation's progress not by activity but by outcomes that reached the people you serve, how far along would you actually be?
The Coordination Trap
Here is why the fog persists. Organisations were designed for humans coordinating with humans.
Every meeting, every approval chain, every reporting structure, every project management ceremony exists because humans building things together need to synchronise. Who is doing what. When it is due. How it connects. These were real problems, and coordination was the real solution.
But when LLMs and agents change the production layer, when a lot of what used to take teams and weeks can be produced by a few people and a model in hours, those coordination structures don't become optional. They become friction. The meetings still happen. The roles still exist. The processes still run. But they are no longer solving the problem they were designed for. They are solving a problem that has already been solved by the technology underneath them.
The bottleneck has moved.
It used to be production: getting enough skilled people to build the thing. Now it is articulation: defining what the thing should be with enough precision that it is worth building at all.
This is the shift from coordination to articulation. Organisations optimised for the old bottleneck are stuck. Not because they lack tools. Because their entire operating model is designed around a constraint that no longer exists. They are running a coordination-intensive system in an articulation-intensive world.
And to be clear: some coordination is irreducible. Safety, compliance, complex dependencies, there are reasons decisions must be shared and checked. The point isn't "no meetings." The point is that coordination must cash out into decisions, and decisions must cash out into falsifiable commitments. Otherwise it's just theatre with minutes.
Most organisations cannot simply delete the coordination layer. That is the brownfield reality. People's jobs depend on it. Identities are built around it. Careers have been built around managing that complexity. For many people, the coordination layer is not just their job. It is their professional identity.
This is why the fog is so comfortable. Leaving it has a visible, immediate cost. Staying in it just has a slow, invisible one.
And here's where this gets sneaky: the fog doesn't just live in operations. It lives in anything where outcomes are hard to specify.
Nowhere is that more convenient than "developing people." Because capability is messy to measure. Behaviour change takes time. Context matters. So organisations fall back to what they can count: workshops delivered, modules completed, check-ins scheduled, satisfaction scores collected.
It's the same pattern as the rest of the fog. Activity stands in for intent. Coordination stands in for progress. And everyone gets to feel responsible without anyone having to be falsifiable.
Information Isn't Capability
This is where leadership development, and training in general, quietly goes wrong.
Most of it is built around communicating information.
Frameworks. Models. A workshop. A deck. A nice phrase you can repeat in the next meeting. Maybe a handful of prompts and a check-in with the executive team every couple of months.
That can feel useful. People nod. They take notes. They leave with language for things they already half-knew.
But it's not capability. It's information transfer.
Capability is what shows up on a Wednesday when the pressure is on. When the calendar is full, the incentives are misaligned, and the old habits are waiting like a comfortable chair. That's when you find out what actually changed.
This is why so many learning experiences "work" in the room and fail in the organisation. They benefit the learner in the short term, because the experience is real, but what was learned gets swallowed by a system not designed to support it. The learner goes back to the same environment: the same manager cues, the same performance measures, the same cultural gravity. The environment wins.
And here's the uncomfortable question: if all you're doing is communicating information, why wouldn't a customer just get it from ChatGPT?
Too many providers are still running the old model, information delivery with a thin layer of coaching, because it's easy to package and easy to sell. And when outcomes aren't measured in the work, the only thing you can reliably count on is paperwork: completions, satisfaction scores, renewal conversations, and yes, an invoice.
That's the vendor model: sell knowledge, measure satisfaction, repeat.
But the customer doesn't need more knowledge. The customer needs change that survives contact with reality.
Why I Left
I was in the fog with everyone else. For years. In industries that had lost the thread of who they were serving and why. Surrounded by talented people doing real work inside systems that had quietly detached from their original purpose.
I knew it. I always knew it. That is the part I want to be honest about. I was never confused about what I was building or why. The fog was not mine. It was the context. The default operating mode. The water we were all swimming in.
And like most people in that situation, I told myself the story that it was just how things work. That the gap between what organisations claimed to do and what they actually did was structural, not a choice. That the coordination overhead, the politics, the slow drift away from purpose, was the price of participating in something larger than yourself.
It was not a price. It was a tax on clarity. And I chose to stop paying it.
Not everyone around me saw it that way. Some thought the fog was just the cost of doing business. Some were comfortable enough not to question it. Some questioned it privately but could not see what they would do instead. I understand all of those positions. I held some of them myself, for longer than I would like to admit.
But there is a moment where understanding becomes insufficient. Where knowing that the system is broken and continuing to participate in it stops being pragmatic and starts being a choice.
My choice was to leave.
The decision was not a revelation. Not an epiphany. It was simpler and harder than that. I looked at the gap between what I believed was possible and what I was spending my days doing, and I decided to stop pretending for a paycheque.
What Building Without Fog Looks Like
It starts agent-first, not bolted onto an old workflow.
The point isn't "AI everywhere." The point is using AI to lift human capability from the start, not just automate the process around it.
So every function is designed around that assumption. You start upstream by writing specifications. Not "deliver a programme," but: what perspective is being built, what behaviour should change, and what evidence would prove it transferred into real work.
Then make that evidence hard to avoid. Give every goal two numbers: the status the leader sets, and the health the system calculates. A leader can mark something "in progress" for months. The system doesn't care. It compares time elapsed against actual progress and flags the gap. If you're 80% through the timeline and 30% done, you're stuck, regardless of what you wrote in the update. That delta between claimed status and computed health is a stuckness metric. It's the fog made visible.
Anchor the rest to what peers and direct reports actually saw, not what someone reported about themselves. None of it should rely on satisfaction scores. All of it should be falsifiable.
Machines handle the production and orchestration between intent and evaluation. Humans do the upstream work that cannot be delegated: defining intent, judging quality, and deciding what to change when reality disagrees with the plan.
This is the crucial difference: most training is designed as a delivery event. Design for experience in the flow of work instead, because that is where behaviour is shaped, reinforced, and either kept or lost.
If the learner returns to a system that doesn't support the new capability, it won't stick. So the system has to be part of the design: prompts at decision moments, practice under real constraints, reflection when consequences are fresh, reinforcement where managers actually influence behaviour.
If the only thing that changes is what people know, it failed. The target is what they do, in the environment that used to pull them back to default.
Here's one ugly detail: agents don't politely ask for clarity. If your specs are vague, they produce plausible nonsense at scale. Agent-first forces discipline. You spend more time on evaluation than you expect. You need tight definitions, hard gates, and someone willing to say "this didn't work" even when the output looks busy and impressive.
That's the trade.
Less coordination theatre. More articulation. Less content delivery. More capability transfer. Less "we ran the programme." More "did the work change?"
Most established organisations can get there. But it's a migration, not an upgrade. The distance between where they are and where they would need to be is not a technology gap. The technology is available to everyone. The harder question is whether an organisation will commit to falsifiable outcomes, then design its systems, human and machine, to make those outcomes real.
The Question
The fog does not lift because you buy better tools. It does not lift because you attend another conference, run another pilot, or hire another head of AI strategy.
The fog lifts because someone decides it is not acceptable.
That's the gap I keep coming back to. Not a technology gap. A willingness gap. The distance between frontier organisations and everyone else is measured in a question most organisations have never been forced to answer:
What, precisely, are you committed to, and would you know if you were failing?
And if your answer includes "developing people," make it sharper:
What would you be able to point to, in the work itself, that proves the capability stuck?
AI has made production cheap. Drafts, variations, summaries, scenarios, even first-pass plans can be produced in hours. That doesn't solve the real constraint. It exposes it.
The scarce input now is intent. Clear outcomes. Testable definitions of "better." If you can't specify what success looks like before you start, you just get faster output pointed in the same fuzzy direction.
The fog is how it has worked. Past tense.
Will you stop accepting it as normal?
