---
title: "What Should a CEO Actually Do? In-Source the Strategy"
description: "What should a CEO actually do when AI runs execution? Own the one thing that cannot be delegated: the connected model of how the whole business fits together."
url: https://buildfirstbrain.com/journal/outsourcing-execution-in-sourcing-strategy/
canonical: https://buildfirstbrain.com/journal/outsourcing-execution-in-sourcing-strategy/
author: "Lawrence Arya"
authorUrl: https://www.linkedin.com/in/vibecoding/
published: 2026-06-09
updated: 2026-06-09
category: "AI & Cognition"
tags: ["ceo", "solopreneur", "strategy", "first brain", "ai-cognition"]
lang: en
---

# What Should a CEO Actually Do? In-Source the Strategy

> **TL;DR** What should a CEO actually do? When AI and a small team can run execution, the leader's real job narrows to one thing that cannot be outsourced: maintaining the connected model of how the whole business fits together, the strategy, the trade-offs, and the reasoning behind each decision. Execution is increasingly cheap to delegate; judgment about what to build and why is not. The Build First Brain approach is the discipline behind that job, treating the organization as a knowledge graph the leader keeps coherent.

What should a CEO actually do? Increasingly, one thing: hold the connected model of how the whole business fits together, and protect the judgment behind each major decision. As AI agents and a small team absorb more of the execution, the parts that used to fill a leader's week, drafting, coordinating, reporting, become cheap to delegate. What does not delegate is the strategy: knowing what to build, what to refuse, and why each call connects to the others. The job is shifting from doing the work to owning the map of the work. Outsource execution, in-source strategy, and treat the company as one coherent model you keep in your head rather than a pile of tasks you personally run.

## What should a CEO actually do?

The core job is to set direction and allocate the company's scarce resources toward it, and that has quietly become almost the whole job. A [chief executive officer](https://en.wikipedia.org/wiki/Chief_executive_officer) has always been responsible for major decisions and overall direction, but for years that sat buried under operational work that a founder also had to do by hand. As tools take over the operational layer, the residue is the part that was always most valuable: [strategic management](https://en.wikipedia.org/wiki/Strategic_management), the work of deciding where the business goes and which trade-offs it accepts to get there.

This is not a claim that execution stops mattering. It is a claim about where the leader's time should go. When a small number of people plus AI can ship the work, the binding constraint is no longer hands, it is judgment: the clarity to choose the right direction and the coherence to keep a hundred small decisions pointing the same way. That coherence lives in one mind, which is why the founder's own thinking becomes the bottleneck, the argument in [how a one-person company is limited by the founder's mind](/journal/the-cognitive-architecture-of-the-1-person-unicorn/).

## Why execution is the part you outsource

Execution is delegable because it can be specified, checked, and repeated, which is exactly what makes it cheap. Once a task has a clear definition of done, it can be handed to a person, a contractor, or an AI agent, and verified against that definition. This is just [outsourcing](https://en.wikipedia.org/wiki/Outsourcing) extended to cognitive work, and it follows the old logic of the [division of labour](https://en.wikipedia.org/wiki/Division_of_labour): specialized, well-defined work gets done faster and cheaper by whoever or whatever is built for it.

The leader's mistake is to keep doing this work because they are good at it. Being the best writer, coder, or operator in the company is no longer an advantage once those outputs can be generated and checked at scale, a shift covered in [what to do when AI can do your job](/journal/you-automated-your-job-now-what/). Holding onto execution feels productive and quietly starves the one job nobody else can do. The skill is to make the work specifiable enough to hand off, then step back to the layer above it.

## What in-sourcing strategy actually means

In-sourcing strategy means keeping the connected reasoning of the business inside the leader's own mind, not just the decisions but the why behind them. It is the difference between a leader who can list this quarter's priorities and one who can explain how each priority depends on the others, what breaks if one is dropped, and which assumption the whole plan rests on. That web of dependencies is the strategy, and it cannot be handed to an agent because it is not a task, it is a model.

Picture a founder running a software product with two people and a set of AI agents. The old version of that founder spent the week writing code, answering tickets, and drafting copy. The new version spends it deciding which single feature actually moves the business, holding why that feature beats the ten alternatives, and keeping the pricing, the positioning, and the roadmap consistent with each other. The agents write the code; the founder holds the reason it is worth writing. If that founder gets hit by inspiration in the shower, it is a connection between two parts of the business, not a faster way to close a ticket. That is in-sourced strategy in practice: the output is coherence, not artifacts.

This maps directly onto a [core competency](https://en.wikipedia.org/wiki/Core_competency): the small set of capabilities a company should never outsource because they are the source of its advantage. For the modern leader, the core competency is the coherent model itself. Keep that in-house, in your head, and you can direct any amount of outsourced execution. Lose it, and you become a manager of disconnected outputs with no way to tell whether they add up. **The work that cannot be delegated is holding the whole picture, so the leader's real job is to be the one place the business stays coherent.**

## Execution versus strategy, by what can be delegated

The useful split is not by importance but by what can be specified and handed off. Set the two layers side by side and the leader's shrinking-but-deepening job comes into focus.

| Dimension | Execution layer | Strategy layer |
| --- | --- | --- |
| Can it be specified | Yes, with a clear definition of done | No, it is a living model |
| Who or what does it | Team, contractors, AI agents | The leader, in their own mind |
| How it is checked | Against the spec | Against reality and coherence |
| Cost trend | Falling fast | Rising in relative value |
| Effect of losing it | Replaceable | The business loses direction |

The right-hand column is what survives as everything else gets cheaper. A leader who spends the week in the left column is optimizing the part the market is busy commoditizing.

## How to manage AI agents without becoming their clerk

The leader's role over AI agents is to be the verifier and the source of intent, not another worker in the queue. Agents can run tasks, but errors compound silently when one feeds the next, so you can only safely run as many as you can actually check, the constraint detailed in [how to manage a swarm of AI agents](/journal/the-ceo-of-the-swarm-managing-ai-agents-natively/). The real advantage comes from defining the work precisely at the top and verifying it at the bottom, while the machine fills the middle.

The highest-return version of this is to perfect the root of a problem and let the branches generate themselves, the idea behind [defining the root node and letting AI build the leaves](/journal/the-leverage-of-the-root-node/). None of it works if no human holds the cross-domain picture, because no agent resolves a contradiction between two domains it cannot both see, which is [why AI cannot truly manage other AI](/journal/ai-middle-management-is-a-myth/). The leader supplies the coherence the swarm structurally lacks. This is the order the Build First Brain approach insists on: build the connected internal model first, then let agents execute against it.

## Where the leader still has to touch execution

The clean split between strategy and execution is a direction, not an absolute, and pretending otherwise is a trap. Early in a company, before the work is well understood, the leader often has to do the execution precisely to learn what the strategy should be, because the model is built by contact with the real work. Founders who delegate too early outsource the very experience that would have taught them the business.

There are also decisions where the detail is the strategy: a key hire, the core of the product, a make-or-break customer relationship. Here, staying close to execution is not a failure of delegation, it is judgment about where the leader's attention actually matters. The discipline is to know the difference, to hand off the work that has become specifiable and to stay hands-on where the doing still teaches you something only you need to know. A useful test is to ask whether finishing this task yourself would change your model of the business; if it would, do it, and if it would only produce another deliverable, delegate it.

## Key takeaways: the modern leader's real job

As execution gets cheap to delegate, the leader's job narrows to the one thing that cannot be outsourced: the coherent model of the whole business. A few points to carry:

- Outsource execution once it can be specified and checked; that is what makes it cheap.
- In-source strategy, meaning the connected reasoning and trade-offs, because it is a model, not a task.
- Over AI agents, be the verifier and the source of intent, not another worker in the queue.
- Do the execution yourself early, when doing the work is how you learn what the strategy should be.
- Stay hands-on where the detail is the strategy, like a core hire or the heart of the product.

The most useful shift is to stop measuring a week by output produced and start measuring it by how coherent the whole picture has become, since coherence is the asset that compounds while everything else gets cheaper. The book [Building Your First Brain](/journal/cognitive-mapping-how-to-build-your-first-brain/) is free for the first 1,000 readers and goes deeper into building the connected model that the modern leader's job now depends on.

## Frequently asked questions

### What should a CEO actually do?

Set the direction and hold the connected model of how the whole business fits together, including the reasoning and trade-offs behind each major decision. As AI and a small team absorb execution, this strategic layer is what cannot be delegated, because it is a living model rather than a specifiable task. The practical job is to make execution specifiable enough to hand off, verify what comes back, and keep the whole picture coherent in one mind.

### Should a founder outsource execution to AI?

Yes, once the work can be defined and checked, because a clear definition of done is exactly what makes a task safe to delegate to a person or an AI agent. The mistake is holding onto execution because you are good at it, which starves the one job nobody else can do. Keep the strategy and the cross-domain judgment in-house, and push the specifiable work outward. The leader's edge is verification and intent, not personal output.

### What does it mean to in-source strategy?

It means keeping the connected reasoning of the business inside your own mind, not just the list of decisions but the web of dependencies behind them. A leader who in-sources strategy can explain how each priority relies on the others, what breaks if one is dropped, and which assumption the plan rests on. That model is the company's core competency and the source of its direction, so it should never be handed to an agent or buried under operational busywork.

### Can AI run a company without a CEO?

Not coherently, because AI agents execute tasks but do not hold the cross-domain picture that resolves contradictions between parts of the business. Errors also compound silently across a chain of agents, so someone has to verify and supply intent. The human role shrinks in volume and grows in importance: fewer hours doing the work, more weight on the judgment that keeps a hundred decisions pointing the same way. The leader becomes the coherence the machine lacks.

### When should a leader still do the work themselves?

Early, when doing the execution is how you learn what the strategy should be, and in the few places where the detail itself is the strategy, like a critical hire, the core of the product, or a make-or-break relationship. Delegating these too soon outsources the experience that builds your model of the business. The discipline is to hand off work once it is specifiable and stay close where the doing still teaches you something only you need to know.

## Dive deeper in

- [How a one-person company is limited by the founder's mind](/journal/the-cognitive-architecture-of-the-1-person-unicorn/)
- [How to manage a swarm of AI agents without drowning](/journal/the-ceo-of-the-swarm-managing-ai-agents-natively/)
- [What to do when AI can do your job](/journal/you-automated-your-job-now-what/)
- [The advantage of perfecting the root node](/journal/the-leverage-of-the-root-node/)

---

Source: https://buildfirstbrain.com/journal/outsourcing-execution-in-sourcing-strategy/
Author: Lawrence Arya — https://www.linkedin.com/in/vibecoding/