The AI does the work.
A human makes the calls.
The Sunday Briefing is a weekly email for service design professionals: market trends, salary benchmarks, and notable roles. This page shows exactly who makes each part of it.
The Scale
Who does what
Five stages, one owner each: the AI or a human. The bar below lays out the running order, and the cards name the stages.
Analysis
AIEvery number, computed.
Angles
AIStory options, proposed from the numbers.
Selection
HUMANOne angle, chosen.
Draft
AIThe article, written.
Review
HUMANThe final edit.
Three of the five stages run without a human. The two judgment stages, choosing the angle and the final edit, do not.
The Numbers
Computed, not composed
Every statistic in the briefing is computed, not typed. Before an article exists, the AI writes analysis code, the code runs against the week's data, and the numbers come out the other end. The same data in, the same numbers out.
weekly data -> analysis code -> numbers
The Workflow
Five stages, two owners
01
AIAnalysis
The AI writes analysis code and runs it against the week's job data. Every count, median, and share that the briefing cites comes out of that code. Run it again on the same data, and the same numbers come back.
02
AIAngles
From those numbers, the AI proposes angles for the week's article: what moved, what is worth a reader's time, and what the data can support. Each proposal travels with the figures behind it.
03
HUMANSelection
A human reads the proposed angles and picks the one the briefing will pursue. This is the editorial call of the week, and it belongs to a person.
04
AIDraft
The AI writes the full article around the selected angle. The computed numbers are its working material, and they reach the draft exactly as they came out of stage one.
05
HUMANReview
A human reads the draft against the numbers and edits. The pass is usually light: copy edits, a cut, a clearer sentence.
Why This Split
Volume for the machine, judgment for the human
One person cannot crunch a week of market data by hand, and a language model should not be trusted to invent statistics. That is the split in one sentence: the machine takes the volume, a human takes the judgment.
The split has limits. A draft can still misstate what the numbers say, which is what the review pass checks for. Where a figure carries a caveat, a small sample or a gap in the data, the caveat travels with the figure, and the disclaimer covers the rest.
Elsewhere
Also on this site
A similar machine pipeline runs behind the X-Ray: language models read every job listing in a nightly batch and fill in fields like skills, tools, seniority, and salary. How the X-Ray works covers it in detail.
Get It Weekly
Get the briefing
The Sunday Briefing arrives weekly: market trends, salary benchmarks, and notable roles from the service design job market, in your inbox.
Subscribe Free