Author: David Aragon

Everyone has an opinion about AI coaching. Most of those opinions are based on interactions with tools that aren’t actually doing coaching — they’re doing cheerleading.

There’s a meaningful difference. Cheerleading validates. It encourages. It finds the positive angle. It tells you you’re doing great and here are some things to consider. It feels good and produces almost no change.

Coaching challenges. It finds the gap between where you are and where you’re trying to go. It names what’s in the way — including the things you’re doing to yourself. It asks the question you’re avoiding. It doesn’t let you off the hook with a good-sounding explanation.

For AI to do the second thing, it needs enough context to know where you’re actually trying to go. It needs your real goals, not a generic version of success. It needs your history — the patterns in what you’ve tried, what’s worked, what hasn’t. It needs to know your excuses, so it can recognize them when they appear.

Without that context, AI coaching is structurally limited to cheerleading. It can give you frameworks and encouragement. It can’t hold your specific pattern up to you and say ‘here’s what you always do at this point, and here’s why it doesn’t serve you.’

That’s the difference between a tool that makes you feel better and a tool that makes you better. The first is valuable in its way. The second is the thing worth building toward.

The standard for AI coaching shouldn’t be ‘is it helpful?’ It should be ‘does it make me better?’ Those produce very different products.

Every time you move between different types of work — strategic thinking to operational execution to communication to analysis — there’s a transition cost. You know this intuitively. What you probably underestimate is how large that cost actually is.

Research on cognitive switching suggests the cost isn’t just the seconds it takes to reorient. It’s the depth of thinking you lose. Deep work — the kind that produces real insight, real progress on complex problems — requires sustained focus over extended periods. Every switch resets the depth counter. You have to rebuild concentration before you can produce anything at the highest level.

For most operators, the day is a constant sequence of context switches. Email, then a meeting, then a strategic question, then another meeting, then operational details. The calendar is full. The output is thin.

The operators who produce disproportionate results tend to solve this the same way: they protect blocks of uninterrupted time for their highest-value work. Not because they’re better at focus by nature — but because they’ve made structural decisions that create the conditions for focus.

This is a planning problem before it’s a willpower problem. If your day is structured in a way that makes deep work impossible, discipline won’t save you. The structure has to change first.

Your most important work deserves your best thinking. Your best thinking requires uninterrupted time. Uninterrupted time requires a plan that protects it. That plan has to be made before the day starts, when you’re not yet in the flow of requests and interruptions.

Protecting your deep work time isn’t a luxury. It’s the highest-leverage decision you make every day.

There’s a critical distinction that most decision-making frameworks ignore: the difference between actually making a decision and constructing a justification for a conclusion you’ve already reached.

Most people, most of the time, are doing the second thing. The decision is made emotionally or intuitively, sometimes before the analysis even begins. What follows isn’t reasoning — it’s rationalization. The research confirms what you already believed. The analysis supports the conclusion you were already leaning toward. The advisors you consult are the ones most likely to agree.

This isn’t a character flaw. It’s how human cognition works under conditions of excitement, fear, or time pressure.

The way to catch it is to pay attention to how you respond to disconfirming information. If you’re genuinely deciding, disconfirming information changes your position or at least creates real uncertainty. If you’re rationalizing, disconfirming information gets dismissed, explained away, or simply not sought.

A useful test: before you finalize any significant decision, deliberately seek out the strongest case against it. Not a strawman — the real, best version of the argument for not doing this. If you can’t articulate it, you haven’t thought about it enough. If you can articulate it and it doesn’t change your thinking at all, ask yourself why.

The goal isn’t to be paralyzed by doubt. It’s to make sure the decision you’re making is actually a decision — not a conclusion you arrived at before the analysis started.

The operators who make the best decisions over time aren’t the ones who are always right. They’re the ones who can tell the difference between deciding and justifying.

If you’ve used AI for anything consequential, you’ve noticed the hedge. The careful qualification. The ‘on one hand, on the other hand.’ The conclusion that sounds authoritative until you read it closely and realize it’s saying almost nothing.

This isn’t a bug. It’s a design choice.

AI systems are trained to minimize the risk of being wrong. The safest output is always the one that covers all possibilities, qualifies every claim, and presents multiple perspectives without committing to any of them. That output is very hard to criticize. It’s also not very useful.

The operators I’ve worked with over the years don’t need more perspectives. They’ve usually considered the obvious angles already. What they need is pressure-testing — someone or something that will engage with their specific position and find the weaknesses in it.

Pressure-testing requires a point of view. It requires being willing to say ‘here’s what’s wrong with this thinking’ rather than ‘here are some considerations you might want to weigh.’ It requires the advisor to actually engage with your thesis rather than present a balanced overview of the issue.

Generic AI almost never does this. It’s been trained out of it.

The AI that’s useful for high-stakes decision-making is the one that will tell you what it actually thinks given your specific situation and framework — not what a reasonable person in your general position might consider. That requires knowing you. It requires having context. And it requires being calibrated to your actual framework rather than to the goal of being agreeable to everyone.

The hedge is a signal. It tells you the tool doesn’t know you well enough to take a position.

Incident command operates under conditions that expose every weakness in a planning system: time pressure, incomplete information, rapidly changing circumstances, high stakes, and teams that need clear direction without lengthy explanation.

In that environment, a tool that requires re-briefing from scratch every time you open it isn’t a tool. It’s a liability.

The ICS framework — Incident Command System — exists precisely because ad-hoc decision-making under crisis conditions produces inconsistent, dangerous outcomes. It builds structure into chaos. It creates shared mental models across a team that may have never worked together. It gives every person in the chain of command a clear understanding of their role and authority.

AI that supports incident command needs to meet the same standard. It needs to know the incident — the objectives, the current operational period, the resources deployed, the constraints in play. It needs to maintain that context across the entire incident, not reset every session. It needs to support the ICS structure, not operate outside it.

The difference between AI that’s generically helpful and AI that actually supports incident command is the difference between a knowledgeable stranger and a briefed operations chief. One requires you to explain your situation from scratch every time you need a decision supported. The other already knows it.

In a crisis, the time you spend re-briefing your tools is time you’re not spending on the incident. That cost is real. For the operators who work in high-stakes environments, it’s unacceptable.

There’s a reason the operators who are serious about fitness tend to be serious about their business metrics. They’ve already learned the same lesson in a different domain: what you measure changes how you behave.

When you start tracking your food, your training, your sleep — not obsessively, but consistently — you discover things you didn’t know about your own patterns. You thought you were eating well. The data says otherwise. You thought you were training hard. The log reveals you’ve missed four sessions in the last two weeks.

The data doesn’t lie and it doesn’t care about your intentions. It only reflects what you actually did.

Business operates on the same principle. Most operators have a story about how their business is doing that’s more optimistic than the numbers warrant. Not because they’re dishonest — because they’re human. We remember the wins more vividly than the losses. We weight recent positive signals heavily. We find explanations for bad data that protect our thesis.

The practice of tracking — whether it’s your macros or your conversion rates — is fundamentally a practice of confronting reality as it is rather than as you hope it is.

The operators who do this consistently across both domains tend to compound faster. Not because they’re more talented. Because they’re operating on more accurate information about themselves.

Measurement is not about control. It’s about clarity. And clarity, sustained over time, is one of the most powerful competitive advantages available to any operator.

The monthly review is productivity gospel. Block an afternoon at the end of the month, review what you did, assess progress against goals, plan the next month. It’s in every productivity book written in the last twenty years.

It also doesn’t work for most people. Not because the idea is wrong — but because the cadence is wrong.

A month is too long. By the time you sit down to review, the decisions that shaped your month are already in the past and feel fixed. The drift has already happened. You’re doing archaeology, not navigation.

What works better is a daily practice that takes five minutes, not a monthly practice that takes five hours.

Every morning: what must happen today, what should happen if possible, what can wait. Every evening: did what needed to happen actually happen, and if not, why not?

That simple loop, done consistently, catches drift in real time. You notice on day three that your A-priority hasn’t moved, not on day thirty. You adjust while adjustment is still cheap.

The monthly review becomes useful only when it’s reviewing a month of daily decisions — not trying to reconstruct what happened from memory. Use it to look for patterns across your daily logs, not to substitute for the daily practice.

The granularity of your planning practice should match the granularity of your decisions. If you’re making consequential decisions daily, you need daily accountability. Weekly reviews for people making weekly decisions. Monthly reviews for people operating on monthly cycles.

Most operators make daily decisions but review monthly. That gap is where drift lives.

Most operators have advisors. Very few of them get consistent, actionable value from those relationships.

This isn’t because the advisors are bad. It’s a structural problem with how advisory relationships work.

An advisor who meets with you quarterly can only know what you tell them. They’re working with a compressed, curated version of your situation — the highlights you can communicate in a one-hour meeting, filtered through whatever framing you bring to the conversation that day. They don’t have access to the full texture of how you operate, the context behind your decisions, the history of what you’ve tried.

So they give you advice that’s reasonable given what they know. Which is, by definition, incomplete.

The best advisors solve this partially through relationship depth — they’ve known you long enough that they can read between the lines, they remember the deals you’ve discussed before, they have a mental model of how you think. But that takes years to build and is expensive to maintain.

The other limitation is social. Advisors have relationships with you. That creates friction around the hardest kind of advice — the kind that challenges your thesis when you’re excited, that tells you what you don’t want to hear, that holds your framework against you when you’re trying to bend it.

Good advisors push through that friction. Most don’t, consistently.

The question isn’t whether to have advisors. It’s whether the advice you’re getting is actually calibrated to you, or calibrated to what’s reasonable for a generic operator in your position. Those produce different recommendations. Only one of them is useful.

When people talk about AI alignment, they usually mean the big philosophical question — how do we make sure advanced AI systems pursue goals that are good for humanity?

There’s a smaller, more immediately practical version of the same problem that almost nobody discusses: how do you align the AI you’re using right now with your specific goals, your specific framework, your specific definition of a good outcome?

Generic AI is aligned with the average user. It’s calibrated to be helpful to the broadest possible audience, which means it’s not precisely calibrated to anyone. For most tasks this is fine. For decisions where your specific situation matters — where your history, your constraints, your red lines are the critical inputs — generic alignment is a liability.

An AI that doesn’t know your investment thesis will give you advice that’s reasonable for a generic investor. An AI that doesn’t know your risk tolerance will calibrate recommendations to a statistical average. An AI that doesn’t know what you’ve tried before will sometimes walk you right back into mistakes you’ve already made.

This isn’t a failure of intelligence. It’s a failure of alignment. The tool is doing exactly what it was built to do — be helpful to everyone. The problem is you’re not everyone.

The operators who get the most from AI are the ones who solve the alignment problem at the individual level — not waiting for the AI companies to solve it at the model level. They encode their framework. They maintain context. They treat alignment not as a technical problem but as a practice.

How aligned is the AI you’re using right now with how you actually operate? That’s worth sitting with.

Everyone talks about the big decisions — the fund raise, the acquisition, the pivot, the hire that changed everything. Those are the ones that get written about, analyzed, turned into case studies.

The decisions that actually shape most operators’ trajectories aren’t the big ones. They’re the small ones made consistently over time.

Which information sources you trust. How much time you spend in meetings versus doing the work. Whether you respond to every message immediately or in batches. How you handle the first sign of underperformance on your team. Whether you re-examine your assumptions when a deal is going well or only when it’s going badly.

None of these feel like high-stakes decisions in the moment. Each one is a small call made under normal conditions. But the pattern of those calls over months and years defines your operating style, your culture, your outcomes.

The problem is that small decisions are almost never examined with the same rigor as big ones. You don’t hold a decision meeting about how you handle your inbox. You don’t bring in advisors to help you think through how you run your Monday morning. You just do what feels right or what you’ve always done.

That’s where drift accumulates. Not in the moments of obvious high stakes — you’re careful there. In the ordinary moments where you’re operating on autopilot.

The discipline that matters most isn’t how you handle the crisis. It’s the quality of attention you bring to the decisions that don’t feel like decisions at all.