A delivery slowdown can have many causes. The Lean lens turns a vague complaint into questions about value, flow, pull, quality, and learning. First locate the slow part of the delivery loop: intent to code, code to build and test, build and test to running software, or production feedback back to intent. section 1.2
For a test pipeline that takes 45 minutes, fails intermittently, and needs manual reruns, ask:
- Value: Which tests prevent defects, and which create false confidence? section 1.2
- Flow: Are changes tested in small increments or a week-sized batch? section 1.2
- Pull: Do commits trigger timely tests, or does testing wait until deployment? section 1.2
- Quality and learning: Are failures addressed and flaky tests investigated, or merely rerun? section 1.2
The answers should change the work, not just label it. Possible moves include parallelizing slow suites, systematically fixing flaky tests, running focused checks on each commit, and making failures block merges rather than deployments. section 1.2
Compare a small change with and without AI
Implement the same small API endpoint or validation rule by hand and with an AI coding assistant. Compare how easily each version can be explained, reviewed, changed, or owned later. Look for assumptions and understanding pushed downstream, even if generation was faster. Use the findings to draft constraints for recurring AI-assisted work. section 3.5 section 3.0
More code or faster production is not necessarily more value. Ask whether work improves delivery or merely adds inventory, rework, or future operating cost—especially when AI makes output cheap to generate. section 2.1 section 3.0
