Where your next senior engineer is going to come from
Why the training work was automated first, why every number you keep says you are fine, and what rebuilding a rung actually costs.
Four questions this volume answers
Skill formed by doing the easy version of hard work next to somebody doing the hard version, and being corrected. That easy work was the most automatable, so it went first, by nobody's choice. Every number you keep measures output, which really is up, most of all for juniors; none measures what someone could do with the tool taken away.
What work reached a first-year person three years ago, and does it still? How many promotions into the middle level per year, the only line that measures the ladder rather than the team? How long do middle-level roles take to fill? And when told “they learn by reviewing the agent's work”: checking an answer and producing one are different skills.
Promotions into the middle level falling while output rises; middle-level roles open for months and filled from outside; first-year people doing review, integration and explaining instead of small pieces of work end to end. Run the afternoon test: take the tool away on a real task and note where the first stall happens.
Whether to pay for the lower rungs: regularly give juniors bounded work an agent could do, make a senior available to correct it quickly, and let the result ship. Name it on the plan with time attached, defend the output dip yourself, and ask for the measurement more than once; a manager below you can't, because they pay and someone else collects. Or decide openly to hire the level from outside.
The work your juniors learned on is the work you automated first.
Skill formed by doing the easy version of hard work, repeatedly, next to somebody doing the hard version. That easy work was the most automatable work in your organization, so it went first — not through anybody's error, but because the properties that make work automatable and the properties that make it teach are the same properties.
This volume is twelve short steps on what that costs, why every instrument you have says you are fine, how to get a measurement before the bill arrives, and what the remedy actually costs. It assumes you have read nothing else in this series.
The distinction the whole volume turns on
| the quantity | what measures it |
|---|---|
| what somebody can produce today, with the tool in front of them | everything you keep |
| what somebody could do with the tool taken away | nothing you keep, anywhere |
Those two moved together for about fifty years, so treating the first as a stand-in for the second was a well-calibrated instinct. It stopped being one recently, quietly, and nothing announced it.
The strongest argument against this volume is true, and it gets its own page before the answer. Your juniors really are producing more, and the gain really is largest for the least experienced people. That is a finding about output, and output is not the quantity at issue.
What you can do afterward
Name the work your juniors learned on three years ago and check whether it still reaches a person. Tell reviewing apart from apprenticeship, and stop counting one as the other. Run an afternoon's test that produces a fact about your own team. And price the remedy honestly enough that it survives a budget review.
The mechanism, the test and the four practices are on one page you can print.