R

Rain Intelligence

The models turn a data feed into forecasts, and the feed is half the product. Signals from dockets, advertising, filings and news are enriched by AI into entities, forecasts, damages estimates, client matches, and practice area and jurisdiction tags. Pattern recognition flags companies at risk before a complaint exists, and the Originate product scores each opportunity against an attorney's expertise. Remove the models and an aggregation and alerting service remains, valuable in itself: the company's own framing is that it finds data nobody else will, and the daily reports and alerts would still carry raw signals. The forecasting and matching that make it more than a news feed are the models' work.

Predictions about named companies are made, and how often they come true is never published. No hit rate appears: how many companies flagged as at risk were sued, how many suits arrived without warning, or how close the damages and settlement estimates came. The site offers commercial figures instead. The example dossier, with its summary, companies at risk and theories of liability, shows no trail back to the signals it rests on.

The Originate examples show match scores such as 98 percent and settlement estimates such as $45 million, with no explanation of how either is calculated. The agreement is candid where the marketing is not: section 2.6 of the terms says AI systems may not always be accurate or error free, and makes users responsible for verifying and cross referencing what the service provides.

The pipeline runs unattended, and the only published control is the person who reads the output. Signals are scanned, forecasts generated, opportunities matched to attorneys and reports delivered each morning to inboxes, Teams channels and firm systems, with zero effort required in the vendor's words. The sample report recommends immediate outreach where companies at risk are existing clients. Nothing describes a confidence threshold below which a forecast is withheld, a human analyst check before a dossier naming a company goes out, or what happens when a prediction proves wrong or a match goes to the wrong attorney.

The terms place verification on the subscriber, a contractual disclaimer rather than a described control. The step from alert to client contact is the lawyer's, and the product is designed to shorten it.

Customer logos include plaintiff practices such as Milberg, Edelson and Morgan & Morgan and defense and full service firms such as Cravath, Davis Polk and DLA Piper. Named partners at Gibbs Law Group and Dechert, among others, give attributed testimonials about origination and early client warning. The headline figures are aggregate and unattributed: more than 150 firms, more than 30 AmLaw 200 firms, more than $200 million in attorney fees generated, a 14 day average lead time and no enterprise cancellations. No method, period or named firm stands behind any of them.

The firm's most sensitive business information goes in, and nothing published says how it is handled. To match opportunities to attorneys, Rain works from a firm's client list, matter history and each lawyer's expertise, and the sample reports flag which companies at risk are the firm's existing clients. The subscription terms govern what the subscriber may do with Rain's materials and make any feedback the firm gives Rain's confidential information.

They contain no confidentiality obligation running the other way, no statement on whether firm data trains or improves the models, and no retention or deletion commitment. Nothing addresses separation between subscribing firms that may compete for the same client or case, the first question for a service rival firms use on the same matters.

The agreement is clear about what the service is not, and it names the user's own obligations. Section 7.11 of the terms states that the online services are not legal advice or a substitute for a professional, that no attorney client relationship arises with Rain, and that the subscriber assumes the risks of relying on the service. It adds that an attorney using it must comply with all court rules, disclosure rules and rules of professional and ethical conduct where they practice.

That clause carries weight because Rain's output is a prompt to contact a company or a potential client before a lawsuit is filed. No treatment is given of the solicitation and advertising rules that govern exactly that contact, which vary by state and are strictest for approaching prospective plaintiffs, or of where the forecasts are designed to be used. Defense side use, warning existing clients of coming litigation, raises fewer of these issues.

Nothing published addresses governance of a system that names companies as likely defendants and estimates what they might pay. There is no responsible AI page or policy, and no one is identified as accountable for the forecasts. Nothing describes what is tested before a model change ships, no back test or evaluation is published, and nothing addresses whether predictions fall unevenly across industries, company sizes or jurisdictions.

That last question carries weight because a forecast that a named company faces a securities or privacy suit circulates among dozens of firms. A systematic skew toward particular sectors or smaller companies would shape who gets sued. The terms' acknowledgment that AI may be inaccurate passes the problem to the subscriber rather than governing it.

The published policy covers the website, not the service. The privacy policy, last updated 4 March 2024, is addressed to visitors of the reports website and to job applicants. It describes administrative, technical and physical safeguards, access on a need to know basis, storage and processing in the United States with standard contractual clauses for European transfers, and retention only as long as necessary. It names AWS and Airtable for hosting and HubSpot for analytics as subprocessors acting on Rain's instructions.

That is a real disclosure for its scope. Nothing equivalent exists for the data a subscribing firm provides, including how it is accessed, which AI providers process it and how a firm would be told of a breach, and the subscription terms do not fill the gap.

Loss is allocated in published terms that cover intellectual property and do not reach the forecasts. Section 4 of the subscription terms, effective 1 October 2023, excludes liability for errors or omissions, for the accuracy or completeness of materials, and for interruption or data loss. It caps aggregate liability at the lesser of actual direct damages or the fees paid in the preceding twelve months. Section 4.5 gives the subscriber an indemnity against third party claims that the services infringe patent, trademark, copyright or trade secret rights, with a right for Rain to procure a license, modify the service or terminate with a pro rata refund.

The risk that a forecast is wrong, that a named company is not about to be sued, or that a damages estimate misleads a partner sits with the firm. The money back guarantee is a commercial remedy rather than recourse for a wrong prediction, and no insurance position is published.

The connections are named, and what travels through each is described. Personalized daily reports go to each attorney's Outlook inbox, urgent signals route to active teams in Microsoft Teams, and a structured prediction database is surfaced inside systems such as Salesforce, iManage and Intapp. That covers the CRM, document management and business development systems large firms already run, and the vendor's pitch is that there are no new tools to learn.

Depth is not published: there is no integrations page, API reference or setup guide, and nothing on what a firm must configure. Nor is there anything on how the firm's own client list, matter history and attorney expertise reach Rain to drive the matching, or which direction data flows on each connection. That inbound path, where the confidentiality questions sit, is the part left undescribed.

Where the service runs and where subscriber data sits are not stated. No tenancy model, hosting region or residency option appears on any product page. The terms address geography only as a licensing matter: section 2.3 makes each subscription specific to a country and bars use outside it for more than 30 continuous days, so the service is sold country by country without saying where it is operated. The website's own data is processed in the United States, but that covers the marketing and reports site rather than the service holding a firm's client relationships.

Whether subscribing firms, some of them competitors, share infrastructure is not stated, and firms outside the United States have no region commitment to rely on.

No independent security attestation is published. There is no security page and no trust center, and no SOC 2 report, ISO 27001 certificate, penetration test or auditor is named anywhere on the site. The only security statement is the privacy policy's general description of safeguards for website data. Nothing on the site offers a route to request documentation.

AI, including generative AI, is confirmed in the agreement, and whose is never said. Section 2.6 of the terms states that some online services use AI, including generative AI for content such as draft emails, contract clauses and summaries. No model, version or provider is named for that drafting or for the forecasting and matching, nothing says where inference runs, and no commitment to notify subscribers of model changes is published.

The privacy policy names hosting and analytics subprocessors for the website, so the company does name suppliers, and no AI provider appears there or anywhere else. The terms mention third party suppliers of materials only as a category. Which company's model sees a firm's client list, if one does, is not answered.

How a firm is committed is published in detail, and no figure appears anywhere. The subscription terms set out access for named authorized users under a services order form that carries the rates, with a transactional pricing plan as an alternative that can be ended on ten days' notice when no order form is in force. They also cover cancellation, a pause of up to three months and refunds, and the site invites qualifying firms to apply for a six month money back guarantee if the service does not deliver substantial return.

The price itself is missing: there is no rate, tier, per seat figure or range for the plaintiff focused Originate product or the enterprise offering for AmLaw firms, and every route ends at a demo.

The audience is set out from three directions, and the limits are not drawn. Plaintiff litigation teams originate class actions and mass torts through Originate, AmLaw partners turn origination into a firm wide process, and business development leaders want every attorney to become a rainmaker. Defense firms use Predict to warn existing clients of litigation coming at them. The subject matter runs from antitrust and securities fraud to consumer privacy and mass torts, and named customers range from solo plaintiff practices to the largest firms.

Which courts, agencies and regulators the signals cover is not stated, nor which practice areas the models do not reach, whether state filings are watched as fully as federal ones, or how far the service works outside the United States.