Building with agents

The System of Growth runs with agents, but it does not hand strategy to them. Agents increase output. People keep judgment.

That split matters because website growth has two different kinds of work: volume work and judgment work.

Agents handle the volume

Agents are good at work that needs speed, structure, and repetition.

They can:

• Research a topic or competitor.
• Classify pages.
• Turn an opportunity into a brief.
• Draft a page from context.
• Compare a page against a quality bar.
• Suggest refreshes.
• Summarize performance changes.
• Find gaps across a portfolio.

That work used to consume the team's time before the team reached the judgment layer. Agents move the team to judgment faster.

People hold the judgment

People still own the calls that can hurt trust if they go wrong.

They decide:

• Whether the strategy is right.
• Whether the claim is true.
• Whether the proof is strong enough.
• Whether the tone sounds like the company.
• Whether a comparison is fair.
• Whether a risk is acceptable.
• Whether the page should ship.

The point is not to remove people. The point is to stop spending human time on mechanical work so people can spend more time on judgment.

Calibration comes before scale

An agent can only work from the context and examples the system gives it. If the context is thin, the agent guesses. If the examples are weak, it copies weak patterns. If the quality bar is vague, review becomes expensive.

Calibration teaches the system what good looks like before volume increases. That means clear context, strong examples, approved claims, reviewer feedback, and a shared definition of quality.

Scale without calibration creates cleanup work. Calibration first makes scale usable.

The operating model is co-operated

The method works best when people and agents run the loop together. Agents handle production tasks across the portfolio. People set priorities, approve proof, refine context, and make the calls that require taste.

That is the operating model: a system where agents increase throughput and people increase quality.

Keep reading