← Back to archive

The 12-Factor Company in the AI Era

27 July 2026 · 59 views · 0 likes · 0 reposts · 237 blog views

AI Startups automation decision-making leadership product

Models are replaceable. Your company’s brain isn’t.

I found an interesting article proposing the idea of a 12-factor company (in the same spirit as 12-factor apps).

  • The company of the past was designed around the bottleneck of human execution.
  • The company of the future will be designed around the bottleneck of human TASTE and JUDGEMENT.

As AI makes execution cheaper and more accessible, the advantage shifts elsewhere:

  • context
  • taste
  • internal knowledge
  • decision-making
  • the ability to learn faster than everyone else

A few useful principles :

  1. Own your context and rent your intelligence.

Models will keep changing. Today’s best model may not be the best one six months from now. Your company’s data/workflows/decisions/institutional knowledge are far more defensible than access to any specific AI provider.

  1. Build your tools and buy your infrastructure

Software is becoming dramatically cheaper to create, so the line between "build" and "buy" is moving.

It makes sense to build the internal tools that reflect how your company actually works, while continuing to buy commodity infrastructure: databases/hosting/cloud.

  1. Try agents before hiring people

For every new repeatable task first ask: can we teach an agent to do this?

Because the value of people is moving toward judgement, creativity, responsibility, taste and not routine execution.

  1. Build small teams with extremely high taste

When execution is abundant, adding more people doesn’t automatically create more value. As it's known small teams can move much faster.

  1. Create rubrics, not just prompts.

For important tasks, agents need more than instructions. They need a clear definition of what "good" looks like: criteria/examples/scorecard they can use to evaluate their own work.

  1. Turn the company into a learning machine.

Store real production traces/failures/capture corrections. Identify repeated patterns then encode that expertise back into your tools, prompts and processes.

The company that learns fastest will outperform the company that simply has the most people or the biggest AI budget.

Your thought and taste are still the advantage. AI should amplify them, not replace them.

A man repairs a printer while a cartoon brain, wearing glasses, relaxes in a chair, commenting on the printer's performance.

Originally posted on X

PS: I'm pretty active on 𝕏 if you'd like to follow along

Related posts

Archive

2026

2025

2024

2018