Software Factory Ladder
A five-level framework for measuring how industrialized an engineering organization's AI-assisted software production is. Proposed by @businessbarista in June 2026.
What It Is
A software factory is an engineering organization where building software works less like craft and more like a production line. The framework places organizations on a 0–4 ladder based on how many steps in the build-ship cycle run without human intervention.
The Five Levels
| Level | Name | Human Role | AI Role |
|---|---|---|---|
| 0 | Artisan | Human writes, reviews, tests, and deploys everything. | None. |
| 1 | Assisted | Human does all steps, but uses AI copilot to write faster. | Glorified autocomplete. |
| 2 | Delegated | Human directs; AI writes the fix and opens the PR. Human still reviews every PR. | Drives execution. |
| 3 | Supervised Factory | AI monitors production, catches bugs, writes fixes, reviews with a second agent, and auto-merges low-risk changes. Humans set guardrails and handle escalations. | Runs the line; humans design it. |
| 4 | Autonomous Factory | AI catches, fixes, tests, and ships before the team knows a bug existed. Humans decide what to build next. | Fully autonomous loop. |
Key Observations
- Most companies believe they are at Level 1 or 2 but are actually still at Level 0 (AI licenses exist but are unused on production code).
- The hardest jump is Level 2 → 3: it requires automated tests you would bet on, an internal platform agents plug into, and written rules defining "low-risk."
- Level 4 is the direction of the frontier, but most organizations should not sprint there because trust, money, and safety decisions still benefit from human judgment.
- The metric "X% of our code is written by AI" is a misleading flex; the real question is how much of the line runs without a human in it.
Why It Matters
The ladder reframes AI adoption from "tool usage" to "process industrialization." It also predicts which functions beyond engineering are next: any function with verifiable output, digital I/O, volume/repeatability, and reversibility is factory-ready. Engineering went first because tests provide automatic ground truth. Functions like QA, data pipelines, and DevOps are close behind. Strategy, design taste, and relationship-building remain least factory-ready because "right" is subjective or only reveals itself far down the road.
Related
- product-trends/saas-vs-agent-native — SaaS vs agent-native disruption
- harness-engineering/overview — Harness engineering for production agent systems
- product-trends/build-then-align-warp — Build-then-align as workflow shift
- product-trends/pm-role-bifurcation-ai-era — PM role changes in the AI era