Darrow

What it is

Darrow reads public data — regulatory filings, court dockets, agency complaints, corporate disclosures, incident reports — and flags where a company’s conduct has probably already created a claim, before any plaintiff has walked into a law firm. Evyatar Ben Artzi (CEO) and Gila Hayat (CTO) founded it in June 2020, alongside Elad Spiegelman, who has since left. The company runs roughly 170 people out of Tel Aviv, with offices in New York and Miami.

The position matters more than the feature list. Almost every legal AI tool in this catalog starts after a matter exists: EvenUp writes the demand package, Supio builds the medical chronology, Filevine runs the case file. Darrow sits one step earlier, on the question of which case the firm should be working on at all. Ben Artzi’s own framing is that Darrow operates upstream, detecting violations, while most legal tech operates downstream on operational efficiency once a case already exists.

The mechanism is a mapped taxonomy, not a general model pointed at the news. Darrow’s legal intelligence team has codified 164 compliance weaknesses spanning securities fraud, antitrust, consumer protection, environmental law, privacy, labor and employment, and ERISA, and tests companies against that set continuously.

What you actually get

On 12 May 2026 the vendor folded its point tools into a single platform with four parts:

• Case Discovery — agents surface candidate exposures, vetted by former AmLaw 100 attorneys before a firm sees them. Each opportunity carries an estimated class size, damages estimate, projected attorney fees, settlement timeline, and defendant revenue.
• Case Evaluation — a case memo with the underlying evidence and procedural documents, plus comparable-case analytics drawn from federal and state courts.
• Portfolio Management — a dashboard tracking settlement value, projected net proceeds to the firm, litigation stage, plaintiff intake, and document collection across the full docket.
• Embedded Intelligence — a chat layer for asking about case merits, defendant history, valuation, and precedent.

The older product names still describe the machinery underneath: Torch, a browser extension that flags violations as you read; PlaintiffLink, which sources and vets actual plaintiffs and routes them to intake; and an underwriting module sold to litigation funders and insurers rather than to firms.

A worked example of the scope: Darrow says it identified $10.3 billion of ERISA exposure over the past year, affecting more than one million plan participants, and that $7.7 billion of it — over 75% — became active legal cases within twelve months. That conversion rate is the number to interrogate in a pilot, because it is the entire thesis.

Pricing reality

There is no published price, no tier list, and no self-serve signup. Every deal starts with a demo.

The structure is hybrid: a subscription scaled to how much coverage a firm wants, plus usage tied to the number of cases, exposures, and scans run. For insurance buyers, usage scales with corporate scans and the number of legal domains included.

The only public numbers come from analyst firm Sacra, which puts the subscription at roughly $500–1,000 per lawyer per month and total annual contracts anywhere from tens of thousands to millions of dollars depending on firm size and case volume. Sacra also describes two charges that are not ordinary SaaS: an exclusivity premium to lock a case away from other firms, and a contingency fee-share routed through Arizona co-counsel arrangements. These are analyst estimates, not vendor-published figures — use the per-seat band as an order of magnitude and get the real quote.

Pin three things in the order form: what exclusivity costs and how long it holds, the scan and exposure entitlement plus the overage rate, and whether any fee-share attaches to cases your own team would have found anyway.

Best for

Plaintiff-side firms with contingency capital to deploy and a real business-development function — mass tort, class action, securities, consumer protection, ERISA — where the binding constraint is finding cases worth taking rather than processing the ones already signed. It also fits litigation funders and insurance underwriters, who buy the same exposure data from the opposite side of the table.

Not for you if your intake queue already exceeds what you can staff. Origination spend on top of a processing bottleneck buys you a longer queue, not more fees. Not for you if you practice on the defense side of a single client relationship, where exposure scanning is a compliance function rather than a pipeline. And not for you if procurement needs published pricing before a demo, because none exists.

Alternatives — and when to pick them instead

• EvenUp — the largest AI platform serving plaintiff firms by valuation, reported above $1 billion, built on demand packages, medical chronologies, and settlement analytics. Pick it when your gap is turning signed cases into demands faster.
• Supio — end-to-end agentic handling of personal injury and mass tort: intake, chronologies, demands, case economics, deposition analysis. Pick it when you want one system for the whole case lifecycle rather than a separate origination layer.
• Eve — the fastest-growing entrant in the segment: $47M Series A in January 2025 led by Andreessen Horowitz, then $103M in September 2025 at a valuation above $1 billion led by Spark Capital, for $164M total. Pick it when you are rebuilding firm operations around AI generally and origination is one item on a longer list.
• Lex Machina and Bloomberg Law — the incumbent litigation analytics layer, and the two names already on most shortlists by installed base. Pick one when you want defensible docket data, judge and court analytics, and case-outcome history. They explain how filed cases behave; Darrow points at a case that has not been filed.
• Mass-tort lead generation agencies and referral networks — the actual status quo Darrow displaces. Pick this when you need signed retainers this quarter and cost per signed case is the metric you are managed on. Darrow sells case theory upstream of that, and the two are complements more often than substitutes.

Watch-outs

• The fee-share structure is a legal ethics question, not a procurement detail. Sacra characterizes Darrow’s contingency-sharing arrangement as conflicting with the fee-sharing rules in force in most U.S. states — ABA Model Rule 5.4 bars sharing legal fees with nonlawyers, and Arizona is the notable jurisdiction that removed that bar in 2021 through its alternative business structure program, which is why the co-counsel routing runs through Arizona. Guard: put any fee-share or co-counsel term in front of your own ethics counsel and your state bar’s rules before signing, and price the pure subscription option separately so you can decline the fee-share and still buy the product.
• Every headline metric is vendor-supplied. The $22 billion in surfaced exposure, the 5 million signals scanned monthly, the 80-plus organizations, the ERISA conversion rate — all of it is Darrow marketing, none of it independently audited. Guard: structure the pilot around cases you already rejected or already ran. Ask the platform to score them blind, then diff its damages estimates and settlement timelines against what actually happened.
• Racing dynamics can erode the value of the signal itself. If several origination platforms surface the same defendant conduct to competing firms, the case gets crowded and its value to any one firm falls. This is a structural risk analysts have flagged for the category, and exclusivity pricing exists precisely because of it. Guard: treat exclusivity terms as a priced feature you negotiate, not a courtesy, and require written scope — which defendants, which claims, for how long.
• The integration story is thin in public. No API documentation is published, and there is no MCP server. Reports of case-management integration describe manual workarounds against legacy systems. Guard: name your case management system — Filevine, Clio, Litify — during the demo and make them show the intake handoff working, not described. If it is a CSV export, plan the headcount to run it.
• The economics assume you can fund the cases you find. A contingency docket sourced faster than it can be financed just moves the constraint to capital. Guard: model the cash requirement of the case volume you expect the platform to produce before you sign, and line up funding capacity in the same quarter.