If Your Work Can Be Micromanaged, AI Will Replace It

I keep hearing it, like a badge, like a moral position.

“We’re a bottom-up company.”

In tech, “top-down” is almost a slur. It smells like bureaucracy, politics, committees, and PowerPoints that never ship.

But the older I get, the less I believe “bottom-up vs top-down” is a management framework.

It is not a framework.

It is a control knob.

And if you treat a knob like a religion, you will eventually burn someone. Or freeze them.

The shower test

Think about the simplest control system you use every day: your shower.

Two knobs, hot and cold.

Nobody says, “I’m a hot-only shower person.” Nobody posts on LinkedIn, “At our home, we are proudly hot-only.” You would laugh, then you would call a plumber.

You mix.

Because the right temperature depends on the context: the season, the country, your mood, your skin, your energy, your day.

Management works the same way. “Bottom-up” and “top-down” are not the goal. They are the hot and cold of the tap.

What you want is the temperature.

And temperature is not universal.

It depends on:

• Culture (country culture, company culture, team culture)
• Project mode (ideation, exploration, delivery, incident response)
• The individual (experience, confidence, personality, appetite for ambiguity)

This is why I don’t believe people are “good managers” in the abstract.

People are good managers in a given situation.

Being a good manager “in general” is not a static skill. It’s a dynamic one: the ability to sense, adapt, and retune without ego.

Bottom-up and top-down are not the framework. They are the plumbing.

Add a third control: pressure

Now extend the shower metaphor a little further, because it gets even more useful.

Most showers have not only temperature, but pressure.

Pressure is cadence, urgency, intensity.

Pressure is the difference between:

• a calm Tuesday exploration
• a “Code Red” delivery sprint
• a production incident
• a make-or-break customer demo

Pressure changes how people behave. It changes attention. It changes tolerance for ambiguity. It changes how much alignment you need, and how quickly you need it.

So, in practice, management is tuning two dimensions:

• Temperature: bottom-up vs top-down mix, decision rights, autonomy
• Pressure: pace, deadlines, escalation paths, focus

A manager’s job is not to “be bottom-up.”

A manager’s job is to create an environment of success, by continuously tuning temperature and pressure to fit the moment.

That was already true before AI.

AI just turned the tap into a firehose.

The moment AI enters, the value function changes

Here is something I hear often in reviews, and it always makes me pause.

An engineer tells me, proudly:

“I found this bug that the machine didn’t find.”

I understand the human impulse behind it. It’s a quiet comparison. A need to prove: “I’m still smarter.”

But as a manager, I don’t care.

Not because bugs don’t matter, they do. I don’t care because the value of an engineer was never meant to be “typing correctness faster than a machine.”

The value is upstream:

• new ideas
• new system designs
• new algorithms
• new product angles
• sharper tradeoffs
• better abstractions
• better orchestration of complexity

Coding, in the literal sense of producing lines of correct text, is increasingly the cheap part.

Orchestrating a machine to produce the right code, for the right intent, with the right tests, with the right interfaces, with the right failure modes, is the new muscle.

That is augmentation.

And if you get augmentation right, something beautiful happens.

AI compresses iteration cycles.

You can go from idea to prototype in hours, not weeks.

You get feedback faster.

You learn faster.

You ship earlier.

You get the reward of impact more often: people using what you built, liking it, trusting it.

Used well, AI can make engineering more joyful, not less.

And yet, in practice, I see the opposite.

A lot.

The “bottom-up resistance” problem

If you default to bottom-up during an AI shift, you often get a predictable outcome.

You get a coalition of thoughtful skeptics, silent resisters, and sometimes outright “unionists against LLMs.”

Not malicious people. Often good engineers.

But the outcome is the same: slow adoption, shallow adoption, performative adoption, or no adoption.

The organization keeps saying “we’re bottom-up,” while the market moves, the product roadmap accelerates, and the gap between what’s possible and what’s delivered becomes embarrassing.

This is where “bottom-up as ideology” becomes dangerous.

Because bottom-up is not management. Bottom-up is one ingredient.

And in an AI transition, many teams need a temporary increase in the “cold” part of the tap.

More top-down.

Not forever. Not as a personality. As a phase.

Because AI is not just a tool you “allow.” It is a capability you must train, the way you train any serious professional skill.

Jobs are not tasks

There is a fear that AI “replaces software engineers.”

I don’t buy it in the simplistic way people say it.

AI replaces tasks. Jobs are bundles of tasks plus accountability, judgment, coordination, taste, risk management, and responsibility.

What changes is the maturity curve.

With AI, the bar for autonomy rises, and it rises faster.

If you stay at the level where your work can be micromanaged, you are in the danger zone.

Because an LLM is better at being micromanaged than you are.

Let me say this bluntly, because it matters.

If your manager can reduce your contribution to a checklist of tiny instructions, they can do that checklist with an AI, faster, cheaper, and with less emotional friction.

That is the micromanagement trap.

And AI makes the trap deeper.

So the goal, for the engineer and for the manager, is the same:

Move your work upstream. Increase your autonomy. Increase your leverage.

The paradox: accept more top-down to earn more bottom-up

Here is the uncomfortable paradox of AI adoption:

Sometimes, you need more top-down now, to get more bottom-up later.

This is exactly how sports training works.

A good coach will make you do workouts you would not choose on your own. They are not “nice.” They can be painful. You might even hate them in the moment.

But the point of structure is freedom later.

A training plan is temporary top-down that produces long-term autonomy.

AI is the same.

When a team is transitioning, “choice” is not always kindness. Sometimes, choice is abandonment.

The manager’s job is to set a few non-negotiables, not to control people, but to protect the future autonomy of the team.

Examples of “healthy top-down” in an AI transition:

• Define what “good” looks like: quality, security, testing, reproducibility
• Make AI usage explicit: where it is expected, where it is forbidden
• Create shared scaffolding: prompts, templates, golden paths, evaluation habits
• Run workshops, pairing sessions, code reviews that teach the craft of orchestration
• Measure outcomes, not ideology: cycle time, defect rate, clarity, maintainability, customer impact

Top-down here is not dominance.

It is alignment plus training.

It is leadership as a multiplier.

The shower model

If you want a simple mental model, here is mine.

Management is tuning temperature and pressure to create an environment of success.

• Temperature (bottom-up vs top-down mix)
• Pressure (pace, urgency, focus, escalation)

The “right mix” changes with:

• culture
• project mode
• individual maturity
• the speed of the external world

AI changes the external world.

It increases the maximum useful pressure. It shortens feedback loops. It makes iteration cheap. It raises expectations.

So the mix must change.

Not because top-down is “better.” Because the conditions changed.

What I want for my teams

I want engineers who use AI the way elite athletes use training tools.

Proactively. Curiously. Without ego.

Not to prove they can beat the machine at typing.

To move faster upstream. To create more value. To build systems they are proud of. To earn trust and autonomy.

And I want managers who can tune the tap without becoming the tap.

Managers who don’t confuse “bottom-up” with goodness, or “top-down” with evil.

Managers who can say:

“This month, we need more structure, because we are learning a new capability.”

And also say, a month later:

“Now you’ve got it, I’m getting out of your way.”

That is what good management looks like.

Not a slogan.

A continuous retuning.

Because the job was never bottom-up or top-down.

The job is creating an environment of success.

And right now, with AI, the environment is moving faster than our instincts.

So we need better plumbing.