Atomic Unit

Why it exists

The Labor Effectiveness Ratio assumes that labor dollars approximate capacity. Two people at the same salary produce roughly comparable output, so cost stands in for work.

AI broke that assumption. One person with real AI leverage now produces several times the output of an identically paid colleague without it. The ratio goes noisy, because overstaffed, underpriced, and AI-illiterate all look the same on it.

The Atomic Unit fixes the denominator. Estimate how many units a task should take. Measure how many it actually consumed. Output becomes measurable independently of tooling.

Why two hours

Short enough that a task decomposed into units is genuinely understood. Long enough that the overhead of tracking does not exceed the value of the measurement.

Anything much shorter turns into timesheet theater. Anything much longer hides the variation you are trying to see.

Two hours is our standard, not a law. Other teams run 4-hour or 8-hour blocks, and either works. What does not work is a block size that changes between teams or between quarters, because the entire value of the unit is that it is a fixed denominator. Pick one, write it down, and leave it alone.

Lineage

The idea comes from velocity units in software estimation, where teams size work in shared abstract units rather than in hours precisely because hours vary by person. We have adapted it to a fixed 2-hour block and named it the Atomic Unit, but anyone who has estimated in story points or velocity units already understands the mechanism.

What is different here is the application. Software teams use units to forecast delivery. We use them to make labor productivity measurable when salary has stopped being a reliable proxy for output.

How estimates stay honest

Estimates come from a shared, calibrated library, never from the person doing the work. If someone is scored against their own estimate, estimates inflate, and they inflate quietly.

The library is built from observed history and refined over time. That makes it a shared standard rather than a negotiation, which is the only version of this that survives contact with a real team.

What it reveals

Give three people the same three-unit task. One delivers in 20 minutes with AI leverage. One takes 6 hours. One takes 2 days.

Same salary band, same title, three different businesses. That spread is invisible on a payroll report and obvious on a unit report.

Common mistake

Introducing units as a surveillance tool. The unit measures the work, not the worker, and a team that believes otherwise will optimize the log instead of the output. Publish the library, explain the denominator, and report velocity at the team level before the individual level.