Audit the Completion Loop When Shipping Falls Behind Generating

This is the full playbook from “AI Makes Drafts Cheap, So Finishing Is the Constraint.” The book prints a shorter version at the end of that chapter, and this page keeps every step and every example.

Run this audit the next time your organization holds more AI-generated options than it has ever held: architectures, prototypes, migration strategies, roadmap alternatives. Run it when shipping has stopped keeping pace with generating.

1. Map your current loop. List each stage, Generate through Learn, and name the person who is accountable for each stage today. A stage with no clear owner is more than an empty box on an org chart. It means nobody is responsible for matching demand to that stage’s capacity, setting entry criteria, or deciding what gets deferred when volume spikes. An unowned stage is a likely place for the next constraint to form, and finding one should trigger an explicit ownership decision before the pressure arrives.
2. Measure Integration Debt now. Count everything your organization generated this week and has not integrated: proposals, branches, prototypes, drafts. If nobody can produce that number quickly, that is the finding. You’re managing a queue you’ve never measured.
3. Find your actual constraint. Walk the loop stage by stage and ask whether each stage’s capacity grew at roughly the rate Generate’s did. The first stage where the honest answer is no is your real bottleneck today, whatever stage your instinct expected.
4. Give Governance Latency an owner and a target. Write down how long a necessary decision may wait before someone with the authority to close it does so, and start measuring the delay this week. Use the method from “Know Which Decisions to Make Early and Which to Delay”: attach an owner and a trigger to any decision worth naming.
5. Choose the four or five metrics that map to the constraint you found in Step 3, drawing from Completion Velocity, Integration Debt, Governance Latency, Validation Load, Engineering Judgment Ratio, and Portfolio Completion Rate. Show them to your leadership team every month, using the six-to-eight-line scorecard discipline from “Give Executives Real Measures of Engineering Health.”
6. Coach one C-suite peer this month. Pick the executive most likely to mistake AI-generated abundance for automatic value, usually the CEO or the CFO. Have the conversation from the online companion before their next planning cycle, not after a missed board question forces it.

Try this with AI.

“Here is everything my organization completed last quarter. Sort it into differentiated work, which compounded an advantage competitors cannot copy, and commodity work, which any competitor could have shipped. Estimate the share of our finishing capacity each category consumed, and tell me what that ratio says about where our scarcest resource is going.”

Give the audit two full quarters. What you’re looking for is a change in the argument rather than one dramatic fix, and it’s the change every other playbook in this book points toward: the room stops debating how much AI-generated work exists and starts debating whether the organization’s capacity to finish that work is growing at the same rate. That is the only version of the question that matters.

All of the playbooks are listed on the playbooks page.