Context Optimisation

There's a specific kind of frustration that comes from not knowing what success looks like. You're working hard, shipping things, checking boxes, and you still can't tell if you're doing the right work. Every time, the root cause is the same: you don't have the context you need.

The best managers I've ever worked with solved this problem instinctively. They knew what to tell you, when to tell you, and, just as importantly, what not to bother you with. One manager I had used to preface things with, "I'm probably telling you more than I should…" and then let me in on something many managers would have kept to themselves. I was grateful every time. It helped me do my job better and it made me feel valued.

Call it judgement, call it taste, call it whatever you want. In the end, it's context curation. It's the difference between teams that execute and teams that spin.

Most organisations are terrible at this. The fortunate few work for managers who are good at it, but the rest of us waste enormous amounts of time in status meetings, Slack threads and hallway conversations trying to piece together what someone should have told us in the first place. AI doesn't change this problem, but it does make it impossible to ignore.

How Humans Get Context

When humans work with humans, context flows in ways we barely notice. Someone sits in a meeting and absorbs the subtext of a decision. They have a hallway conversation that fills in the gap from last week's email. They read a Slack thread, skim a document and ask a colleague a quick question. When they need to learn something they don't know, they reach out to other humans, read more docs, search intranets and piece together what they need.

Over time, they build a mental model of what's happening and why. They get better at this on their own, not because anyone is carefully curating context for them, but because humans are remarkably good at foraging for what they need. Some take notes, some keep it in their head. Regardless of method, the curation happens organically.

Agents Are Different

Agents work nothing like this. An agent draws from whatever context you give it, period. Too much of the wrong context produces lopsided, confidently wrong answers. Too little context produces work that isn't valuable. Unlike a human, an agent doesn't get better at context-gathering over time. The agent you prompt on year two is just as dependent on what you put in front of it as the agent you prompted on day one. It won't overhear anything. It won't ask around. It won't notice that the goal changed last week unless you tell it.

With agents, you have to be intentional about building systems to capture high-quality context and keep it up-to-date. Vague instructions don't work. You have to be really good at giving them the right context, every single time.

The Human Situation Isn't Good

All of that makes it sound like the human approach to context is the one that works, that our messy, organic, self-curating approach has the advantage. The truth though is that most people in most organisations waste an enormous amount of time gathering context. They sit through countless status meetings not because the meetings are useful, but because nobody has done the work of packaging the information in a more efficient way. They have hallway conversations, send Slack messages and write emails asking questions that someone, somewhere in the organisation, could have answered preemptively.

This isn't a feature of how humans work. It's a tax, one we've normalised because we've never had a realistic alternative. At the individual level, it's frustrating. At the company level it's enormously expensive, yet most leaders don't see it because it doesn't show up on any P&L. Organisations hire product managers and middle managers in large part to do exactly this job: gather context, synthesise it and disseminate it to the people who need it. That isn't a side function of management, it's the core function.

Still, despite layers of people whose job it is to solve this problem, most employees have only a fraction of the context they need to be truly effective. We've spent decades throwing human labour at the context problem and called the result "management." It hasn't worked particularly well. In my own experience as a manager, every time an employee came to me with something that was wrong or inadequate, I could trace it back to a lack of context — or their understanding of it. Every time.

Where Agents Change the Equation

AI can do some things dramatically better. It can read all the documents. It can write status reports in seconds. It can surface patterns across projects that no human would spot. These are real improvements to the coordination machinery.

However, they don't touch the historically serious problem: making sure the right people have the right context at the right time. That problem hasn't improved, and agents need progress there even more acutely than people do. A human will eventually figure out that they're missing something and go find it. An agent won't. It will execute, cleanly and quickly, against a version of reality that no longer exists. The cost of bad context goes from "wasted time" to "wrong decisions made at speed."

The Real Job

Managers have always been responsible for context. Most got away with doing it badly because humans are forgiving — they forage, they ask around, they eventually figure it out. Agents won't cover for you. They'll execute cleanly against whatever context you give them, and if that context is thin, you'll get thin work back, at speed and at scale.

That's the shift. Not that management is being replaced, but that the part of management most leaders quietly neglected — deliberately building and distributing context — is now the part with the highest return. Every hour you invest in it compounds. Better context makes your people faster. It makes your agents useful instead of dangerous. It's the same investment paying off twice.

The managers who double down here will pull ahead. The ones who keep running status meetings and calling it leadership won't.

This is the bet Clarity Forge is built on. We put humans and agents in the same workspace, drawing from the same goals, the same projects, the same organisational context. Context you add benefits everyone. When an agent finishes work, the result becomes context for the next person — human or agent — who needs it. The context tax doesn't disappear, but it stops being paid over and over by every individual and starts compounding for the whole company.

That's what the next decade of management looks like. Start building the muscle now.