Vanta and Drata charge $22k–$80k and $20k–$60k a year respectively for compliance automation. The auditor who signs the SOC 2 — the licensed professional who actually performs the regulated work and carries the liability for it — charges $10k–$50k, once (Comp AI).
That inversion is the whole vertical. The margin has left the people doing the work and settled in the workflow layer sitting on top of them, and it did so because of a structural property that most of this atlas does not share: the supply is licensed, not aggregated.
A middleman can keep 70–80% of an expert network call (Expert networks) because the expert bears no liability for the advice. A middleman cannot keep 70–80% of an attorney's fee, because the attorney is personally exposed, prices accordingly, and in most jurisdictions cannot legally share the fee with a non-lawyer at all. The category fits the atlas pattern superficially and fails it structurally.
Whose budget
Small per customer, universal, and non-discretionary — the best sales cycle in the atlas attached to the worst pricing power.
SOC 2 / ISO. Audit $10k–$50k, compliance platform $20k–$80k/year, implementation $2k–$15k, mandatory annual pen test $5k–$25k. Total first year $30k–$150k+, plus 200–500 internal hours (Comp AI).
Finance and accounting. Pilot: Essentials $99/month, Core from $299/month with a dedicated US bookkeeper, tax packages $750–$5,400/year, R&D credit priced at 20% of the credit received, CFO service from $1,750/month. Fondo: $399/month (list $499), TaxPass $1,450/year (Truewind).
Immigration. Lighthouse: O-1 $10,000, H-1B $4,500, EB-1A/EB-2 NIW $15,000. Alma: O-1 $8,000, H-1B $3,500, green cards $10,000. A ten-petition portfolio runs $98,500 at Lighthouse against $72,500 at Alma (Alma). Legalpad, out of Techstars, pioneered the productised-visa model (Techstars).
Fractional legal. Priori Legal reports its marketplace attorneys at 38% below the NYC industry average, an SF average of $265/hour, and flat fees in 31.2% of engagements against 16% industry-wide — with no published take rate (Priori). Lawtrades has moved upmarket to mid-market and enterprise (Legal Dive).
budget: 2. The spend is mandatory, which compresses the sales cycle to almost nothing, but mandatory spend is also spend the buyer resents and shops hard. Nobody pays a premium for the thing they must buy anyway.
Can you get the supply
Yes, and you cannot expand it.
Those two facts usually do not travel together. CPAs, attorneys and SOC 2 auditors are not scarce in the way an elite jailbreaker or a trained bimanual teleoperator is scarce — you can hire them with money. But you cannot create more of them with money, because each one carries a licence, a regulator and personal liability, and the licensing pipeline runs on a timescale no operator controls.
The consequence shows up in company structure. Lighthouse HQ is software plus document preparation, with the legal work delivered by independent licensed attorneys; Alma is itself a licensed law firm (Alma). That distinction is not branding. It exists to route around unauthorised-practice-of-law rules, and it determines who bears malpractice exposure, who can own the fee, and therefore what margin is available.
supply: 2. Available for money, unexpandable by money, and it prices itself.
That same licensing gate sets the clock on the first invoice. speed: 4. Everything else here is fast: the spend is mandatory so there is no budget to argue for, the deliverable is productised at a published flat fee — $399/month bookkeeping, a $10,000 O-1 petition — and a buyer who needs SOC 2 to close an enterprise deal signs the week they are asked for it. What you cannot compress is the practitioner: before the first invoice you must either attach an independent licensed CPA or attorney, as Lighthouse does, or become the licensed firm yourself, as Alma did. That is a month or two of arranging, not a job advert — which is why this is a 4 and not the 5 that Design, video and content production and Outbound and GTM-as-a-service get for buying supply off a job board.
What the spread looks like
Thin where the work is regulated, and everything interesting happens elsewhere.
Three properties cap the take:
- Liability sits with the practitioner, who prices it in. The middleman cannot keep an expert-network share on a fee whose bearer can be sued for it.
- The output is binary and auditable. You got the SOC 2 or you did not; the visa was approved or it was not. That kills the outcome ambiguity that lets outbound and creative agencies hold margin through a bad quarter.
- The buyer can price-shop precisely, because the deliverable is standardised — hence a public $98,500-versus-$72,500 comparison on the same ten petitions.
So the winners monetise around the licensed work rather than on it. Pilot's real margin is in the $1,750/month CFO tier and the 20%-of-credit R&D contingency, not the $299 bookkeeping. And the platform layer takes the recurring money outright: Vanta reached $300M ARR in April 2026, up 69% year-on-year, on 16,000 customers at roughly $19K ARR each, having become "the de facto solution for three-quarters of YC companies" (Sacra).
A marketplace is valued on the fee it keeps from transactions it does not perform; a professional-services firm is valued on the profit left after paying the people who perform them. The first scales with volume at near-zero marginal cost, the second scales with headcount. Public comparables price that gap as a multiple gap, not a growth gap — which is why the software layer here trades at a software multiple while the audit shop underneath trades like a firm. Anyone building in this vertical is choosing which of those two businesses they are, whether or not they say so. See Marketplace, staffing firm, BPO or agency and What the public market pays for labour.
spread: 1.
Can you hold it
Better than anything else in the function section, and this is the vertical's one genuine strength.
Compliance recurs annually by regulation. Bookkeeping recurs monthly by necessity. An immigration case creates a multi-year relationship — O-1 to extension to green card — with a switching cost measured in re-papering a live petition. Churn is low because the alternative to paying you is not doing without; it is doing the same thing with someone else, at a cost the buyer does not want to incur mid-cycle.
The expansion motion is real and documented: Vanta's ARR per customer rose $17K to $19K while customer count grew from 12,000 to 16,000 (Sacra); Brex derives half its revenue from upsell and cross-sell into existing customers (SaaStr).
The ceiling on that: 62% of seed-funded startups shut down within seven years (SaaStr on Carta). Retention that survives everything except the customer's death is still retention that loses two-thirds of its logos. hold: 4.
What AI does to it
It compresses the human hours and does not touch the licence.
Document preparation, evidence collection, control mapping, bookkeeping reconciliation, first-draft petition assembly — all model-tractable, all currently billable. Lighthouse's structure already anticipates this: software plus document preparation, with attorneys attached only where the licence is legally required. Truewind's framing of the Pilot-versus-Fondo comparison is itself an AI-accounting pitch (Truewind).
The result is not deletion but a squeeze from both ends: automation shrinks the billable middle, while the regulated core keeps its price because a model cannot sign an audit opinion or appear before USCIS. A firm whose revenue is mostly the middle has a problem. A firm whose revenue is the software attach does not. ai: 2, and see What better models do to each layer.
What would kill it
It is already dead as an arbitrage — the question is only how fast the remaining human margin compresses. Vanta at $300M ARR and three-quarters of the YC cohort demonstrates that the value in this vertical accrues to whoever automates the workflow, not to whoever assembles the practitioners. An operator entering now as an aggregator of auditors or attorneys is entering the half of the market that the other half is deliberately shrinking.
Two secondary risks. Regulatory structure — the software-plus-independent-attorneys arrangement exists in a live grey area, and an unauthorised-practice ruling against one operator reprices the model for all of them. Price transparency — public $98,500-versus-$72,500 comparisons of identical work are what a commoditising market looks like from the inside.
Who is already there
| Segment | Who | Published price | Structure |
|---|---|---|---|
| Compliance platform | Vanta, Drata | $22k–$80k and $20k–$60k per year | Software; $300M ARR at Vanta (Apr 2026) |
| Audit | Independent SOC 2 / ISO auditors | $10k–$50k per audit | Licensed, personal liability |
| Accounting | Pilot, Fondo | $99–$299+/mo; CFO from $1,750/mo | Services with software attach |
| Immigration | Lighthouse HQ | O-1 $10,000; H-1B $4,500; NIW $15,000 | Software + independent attorneys |
| Immigration | Alma, Legalpad | O-1 $8,000; H-1B $3,500; green card $10,000 | Licensed law firm |
| Fractional legal | Priori Legal, Lawtrades | SF average $265/hr; 38% below NYC average | Marketplace; take rate unpublished |
Where the record is thin
Priori explicitly does not publish one. Neither does Lawtrades, Lighthouse, Alma, Pilot or Fondo. The claim that the take is structurally capped rests on the shape of the businesses — licensure, liability, fee-splitting rules — and on the observable fact that the recurring money sits with Vanta rather than with the auditor. It does not rest on a single disclosed margin, because none exists.
Two further holes. Nobody has published what an auditor, a fractional GC or an immigration attorney nets on marketplace-sourced work versus direct work, so the leakage rate is unmeasured (Getting cut out). And there is no figure anywhere for what share of the compliance dollar has already moved from human hours to platform subscription — the trendline that would tell you how much human margin is left to compress.