All 15 verticals, one screen

Compare

Six scores, one to five, applied identically to every vertical. They are judgements built on the evidence in each page, not measurements — the reasoning lives on the vertical page and the rubric lives in the framework. Sorted by total, which is a blunt instrument: read the columns, not the sum.

Vertical	BUD	SUP	SPR	HLD	AI	SPD	Σ	Read	Spread
Adversarial evals and red-team crowds	5	5	4	3	5	4	26	build	Effective take on crowd output >90% [UNVERIFIED — inferred, no revenue disclosure]
Forward-deployed engineering	5	4	3	3	4	3	22	watch	33–36% on the marketplace route (Turing, disclosed); undisclosed on the build-it-yourself route
Robotics teleoperation and physical-world data	5	4	3	3	4	2	21	build	40–65% gross on quality teleop; 70–90% on egocentric [UNVERIFIED] — in a segment whose price has collapsed
Expert data for frontier labs	5	3	2	2	4	4	20	crowded	27–33% gross margin (Mercor, leaked); 42.8% published take (Prolific)
Expert networks	4	2	5	4	3	2	20	open	~70–80% — $200/hr expert rate against $800–900/hr client billing
Data-centre labour and site brokerage	5	4	2	3	4	2	20	watch	Conventional construction staffing markups run 30–40%. No AI-native intermediary has published anything
Contingency recruiting marketplaces	3	3	3	2	3	4	18	watch	~25–30% of the fee ≈ 5–7.5% of first-year salary
Outbound and GTM-as-a-service	3	2	4	2	2	5	18	crowded	~55–75% gross margin [UNVERIFIED — inferred from two published rate cards]
Compute and capacity brokerage	5	2	1	2	5	3	18	watch	Low single-digit percentage [UNVERIFIED — SF Compute's fee schedule is not public and its docs page 404s]
Community, campus and events	3	3	2	2	3	4	17	watch	30–50% gross for hackathon agencies [UNVERIFIED]. No ambassador-ops intermediary exists to measure
Paid creators, clipping and UGC ad ops	3	2	2	1	3	5	16	watch	9–18% on marketplaces; undisclosed and materially larger in managed agencies
Voice, speech and low-resource language data	2	3	4	2	2	3	16	watch	60–85% on spec'd low-resource collection [UNVERIFIED — buy-side price is a working assumption, not a source]
Sales talent: AEs, SDRs and the people who close	3	2	3	2	2	4	16	avoid	18–33% of base salary as a fee; $9–14K at SDR, $40–63K at VP Sales
Design, video and content production	2	1	4	2	1	5	15	avoid	70–88% [WEAK — supply-side figures are self-reported]; 55–65% on generous assumptions
Compliance, finance and back office	2	2	1	4	2	4	15	avoid	Low, and structurally capped. The margin sits in software attach and contingency fees, not in the spread on licensed hours

01

What the columns mean

Higher is always better for the operator, including for supply difficulty — a supply that is hard to assemble is the only part of this business a competitor cannot copy over a weekend.

BUD

Budget depth

How much money sits behind the buying decision, and how urgently it must be spent.

SUP

Supply difficulty

How hard the seller side is to assemble. Hard is good: it is the only thing a competitor cannot copy in a weekend.

SPR

Spread

What fraction of the money passing through you, you actually keep.

HLD

Holdability

Whether buyer and seller can cut you out once they have met, and whether one buyer can end you.

AI

AI direction

Whether capable models grow this budget or delete it.

SPD

Speed to first dollar

How long from a standing start to a real invoice.

02

The same table in words

Each vertical's one-line read, in rank order.

Adversarial evals and red-team crowds

A brand-new budget line with no salary to be benchmarked against, supply that cannot be recruited by job ad, and a middleman that monetises the crowd's output as software rather than reselling its hours.

Forward-deployed engineering

Deployment is the acknowledged bottleneck and the budget is real; the open question is whether any of these companies converts field work into a renewing licence before the 58x mark has to be justified.

Robotics teleoperation and physical-world data

The only human-data vertical where physical capital blocks the laptop-marketplace playbook — which is why quality teleop still holds a 40–65% margin while egocentric video went to free in eighteen months.

Expert data for frontier labs

The budget is real, urgent and growing. The margin is a staffing margin, the buyers are two companies wearing five names, and every seat at the top table is taken.

Expert networks

A 70–80% take that survived forty years, three compliance scandals and every disintermediation attempt — because the product was never the expert, it was recruitment speed plus indemnity. Whether the incumbents sell to AI labs is the single most valuable unanswered question in this atlas.

Data-centre labour and site brokerage

The only capacity vertical where the unit is not fungible and not indexed, so a real spread can still live there — but the incumbents are ordinary staffing firms at 30–40% markups and nobody has built the AI-native version.

Contingency recruiting marketplaces

The supply-side wedge is real and the AI tailwind is real; the TAM is a tenth of what the pitch says and there is no independent evidence any of it works.

Outbound and GTM-as-a-service

The spread is genuine at 55–75% and the sale closes in weeks, but every retainer is priced against a salary the buyer can look up, and the supply scarcity that made Clay agencies expensive in 2024 was gone by 2026.

Compute and capacity brokerage

Once a supply becomes fungible and indexed, the middleman stops earning margin per unit and starts earning volume times a tiny fee. That is a different business from every labour vertical in this atlas, and it is worth far more only if you become the venue.

Community, campus and events

The labs run campus programmes in-house because nobody sells the operational layer well; the opportunity is real, the evidence that anyone is capturing it is almost entirely absent.

Paid creators, clipping and UGC ad ops

Real budgets, trivial supply, a 9–18% marketplace take and no hold. The defensible position is the managed one, and that is the position the holdcos are buying.

Voice, speech and low-resource language data

The only defensible slices are consent provenance and long-tail language reach; everything else is a documented race to zero, and the spread that makes the pitch work has never been observed.

Sales talent: AEs, SDRs and the people who close

The contingency fee ladder works at VP Sales and above and almost nowhere else: a 90-day guarantee against a 9-month tenure is a bad trade for whoever is holding it.

Design, video and content production

A 70–88% spread on Glassdoor-grade evidence, earned by operations rather than by scarcity — and the entire pitch is a comparison to a designer's salary that generative tooling is dragging down.

Compliance, finance and back office

Non-discretionary spend with short sales cycles and low churn, and none of it accrues to a middleman: Vanta captures $22k–$80k a year while the auditor who does the regulated work captures $10k–$50k once.