Litify

AI Centrality

How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.

The home page calls Litify the only AI-native platform of action, and the product's own history contradicts the claim: the Taylor King Law case study describes a 2021 deployment whose measured gains came from the analytics engine, phone-tree routing and matter plans, with no AI in the account at all. What sits on top now is substantial rather than decorative. Litify ACE is positioned as the AI nerve centre, eight further named agents run from intake qualification through deposition preparation, and machine learning is embedded in document generation, time capture and line-item invoice review. Remove the models and a working intake, matter, billing and reporting platform remains, which is the shape of a legacy system that added a real AI layer rather than one built around it.

Citation Accuracy and Hallucination Disclosure

Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.

Accuracy is claimed repeatedly and measured nowhere. The Litify AI page asserts that being fully native to the platform ensures greater accuracy than standalone AI tools, and the Damages Assistant is described as reconciling bills against records for total accuracy. No test set, no benchmark, no error rate and no failure mode is published on any surface checked on 31 August 2026. The product does not retrieve primary law, so the citator limb of this axis does not bite; what it does do is draft demand letters and interrogatory responses from a firm's own medical records and matter data, and no grounding mechanism is described for that output. Nothing was located that would let a reader open a drafted assertion and trace it to the source page it came from, which is the specific thing a comparable plaintiff product in this pull does publish.

Autonomy and Oversight Model

What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.

Litify sells autonomy in plain terms and publishes no control structure for it. ACE is described as a proactive, autonomous partner that anticipates risk and executes tasks in the background, the platform page describes autonomous AI agents that execute tasks in the background, and the Demand Agent drafts demand letters automatically as damages approach policy limits while the Discovery Agent auto-drafts interrogatory responses. Against that, the only oversight position located is a philosophical one in the Pillars of Responsible AI Development post of 9 July 2024, which states that AI is not a replacement for legal professionals and is not well positioned to engage in the art of case strategy. That is a statement of intent rather than a mechanism: no review point, no approval step, no threshold at which an agent stops, and nothing on what happens after an agent is wrong was located on the product, platform or AI pages. A stated belief about the division of labour is not the same as a published place where a lawyer signs off.

Operational and Outcome Evidence

Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.

This is the strongest evidence set located in the plaintiff lane. The Taylor King Law study names the firm, its size at 75-plus employees, its states, its COO Emily Merryman, and a 2021 go-live migrating from Needles, then gives a baseline of a 30 per cent intake conversion rate moving to 50 per cent and a 46 per cent rise in average attorney fees. A published baseline is rare in this corpus and it is what makes the figure assessable. The method is described step by step rather than asserted: identifying high-performing intake agents from captured workflow data, routing existing-case calls away from them by phone tree, replacing uniform follow-up with if/then timing logic, and adding an automated quality-of-life questionnaire to surface non-economic damages. Further named studies carry their own figures, including EllisDon at 100 per cent faster reporting, McFarlane Law at a 10 per cent lift in billing collection and Cesar Ornelas at 33 per cent more case production without added headcount. Two limits belong on the record: no study carries a publication date or a measurement window, and not one of the located figures is attributed to an AI capability, so the outcome evidence is for the platform rather than for Litify AI.

Privilege and Confidentiality Posture

How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.

The commitments are specific, readable before signing, and none of them is contractual, because no customer agreement is published anywhere on the property. The Litify AI page states that customer data is never retained or used to train the model, the Pillars post of 9 July 2024 describes single-tenant architecture giving each client a walled-off cloud environment, and the security post states that customer data sits in Salesforce data centres to which Litify is not an authorised administrator, with support access requiring the customer to run a Grant Litify Support Team Access process. It also states that third parties including Litify employees, investors, shareholders and board members are not permitted to access customer information. What is missing is what the band expects to be missing at this level: segregation is described between customers rather than between matters inside a firm, which is the level a plaintiff or defence firm buyer actually needs, and nothing addresses what Anthropic or AWS may retain when a prompt reaches Bedrock. Privilege and work product are not mentioned on any surface checked, and the only published legal document, the privacy policy, expressly excludes data held on behalf of customers.

UPL and Professional Responsibility Posture

Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point.

Searched the home page, the Litify AI page, the platform page, the privacy policy, the CA notice link, the responsible AI post, the security post, the customers section and the Salesforce AppExchange listing on 31 August 2026: no disclaimer, no statement of what the product is and is not, no supervision or competence discussion, no jurisdiction limit and no engagement with ABA Formal Opinion 512 or any state bar guidance was located. There is no terms of service on the property to carry a boilerplate disclaimer, so the record here is a true absence rather than a thin clause. This matters more than it would for a reporting tool, because the shipped agents draft demand letters and interrogatory responses and the platform is sold to firms whose staff include non-lawyer case managers and intake agents. The absence is the finding.

AI Governance and Bias Disclosure

Published governance over model behaviour: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.

Litify publishes a governance post, Our Pillars of Responsible AI Development, dated 9 July 2024 and by-lined to VP of Engineering Andrew Heffernan, which is more than most of this lane offers. Read for governance specifically, most of its substance belongs to other questions: the Claude 3 selection, Amazon Bedrock and single-tenant storage are supply-chain and deployment facts, and the value-first pillar is a philosophy about not replacing lawyering. What remains as governance is a set of principles with nothing attached to them. No accountable owner for AI governance is named, no pre-release testing regime is described, no evaluation results are disclosed, and nothing at all was located on bias or uneven output across matter types, claimant populations or case values. That last gap is worth stating plainly on a product whose Case Value Agent and Policy Agent make value judgements about individual injury claims.

AI Safety and Data Stewardship

Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.

Access control is genuinely well described and the rest of the ground is not covered. The security post of 30 June 2022 states that customer data sits in Salesforce data centres where Litify is not an authorised administrator, that documents are held on AWS behind a strict internal approval process, that the support team must be granted access through a named customer-run process, and that employees, investors, shareholders and board members may not access customer information. Against that single strong element: no retention period, no deletion route, no subprocessor list and no incident or breach-notification practice was located on any surface checked on 31 August 2026. The privacy policy that would ordinarily carry some of this states in its opening that it does not apply to personal information held on behalf of users of the services, which removes matter data from its scope entirely. The access-control material is also four years old and has no current counterpart.

AI Liability and Recourse

What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.

No customer agreement of any kind is published. The escalation ladder was run in full on 31 August 2026: the footer of every rendering page carries only a Privacy Policy and a CA Notice at Collection, a search anchored on agreement language returned no Litify terms, and the Salesforce AppExchange listing carries Salesforce's own Terms of Use rather than the provider's. No master subscription agreement, order form, indemnity, liability cap, warranty on output or insurance position was located, and there is no trust portal or documents page from which one could be requested. This is an absence rather than a retrieval failure: the pages that would hold such a document render normally and simply do not exist. A buyer cannot learn who bears the loss when an agent drafts a wrong demand until they are already in a sales process.

Practice Systems Integration Depth

How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.

The platform page displays roughly thirty integration logos, and the names are the right ones for legal work: iManage, NetDocuments, Outlook, DocuSign, Aderant, Epiq, QuickBooks, Chrometa, Time by Ping, RingCentral, Vonage and Case Status among them. The See All Integrations link on that same page resolves to a bare anchor and goes nowhere, and no integrations directory, developer index or connector documentation was located anywhere else on 31 August 2026. So an implementer can learn that a NetDocuments or iManage connection exists but not what it moves, in which direction, or what a firm must configure. Being a Salesforce managed package does bring genuine API reach, but that is the platform's extensibility rather than a documented Litify integration, and it is credited on the deployment question instead of here.

Deployment Model and Data Residency

Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.

The tenancy model is stated and the location is not. The Pillars post of 9 July 2024 describes single tenant architecture in which each client has a unique cloud environment walled off from other organisations, extended to single tenant storage for the AI features, and the product is delivered as a Salesforce managed package requiring a Salesforce platform licence. Beyond that, nothing: no region list, no data residency options, no distinction between where data is stored and where it is processed, and no statement of which Salesforce or AWS regions a firm's matter data occupies. The privacy policy does say servers are in the United States, but it scopes itself to the marketing site and expressly excludes data held on behalf of customers, so it cannot answer the question for the product. Graded against the bands: the C band requires that neither the tenancy model nor the region be stated, and the tenancy model is stated, so C is not available. What holds the grade below A is that the residency half is wholly absent and the tenancy claim rests on a blog post from July 2024 with no current specification page behind it. Regraded from C on 1 September 2026 for consistency with records publishing one of the two limbs.

Security Certifications and Trust Center

Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.

Litify names its auditor, which most of this lane does not. The security post states that Litify has completed SOC 2 Type I and Type II audits, both performed by Sensiba, and that it undergoes annual independent security audits. What is absent is everything that would let a buyer act on it: no scope, no trust services categories, no coverage period, no report date, no trust portal and no route to request the report short of a demo. The post is dated 30 June 2022 and no current attestation was located, so a reader in 2026 cannot tell whether the annual cadence held. Separately, the Litify AI page claims the product is fully configurable to meet HIPAA, GDPR and SOC II standards, which is a statement about configurability rather than an attestation and should not be read as one.

Model Supply Chain Disclosure

Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.

Litify names its model provider in the marketing hero, which almost nothing else in this pull does: the home page describes the platform as built on the infrastructure of Salesforce, AWS and Anthropic, identified as its closed AI model. The Pillars post of 9 July 2024 goes further and names the model to a version, the Claude 3 family from Anthropic, reached through Amazon Bedrock and run inside closed AWS services so client data is not exposed to the broader internet or to open models. Two things hold this below the top of the axis. No commitment to notify customers when the model or provider changes was located, and the version naming is two generations stale, with the current AI page naming Anthropic and no model at all. The supply chain also extends further than the model pages admit, since the Damages Assistant and Instant Demands are powered by Supio, AI Time Capture by Billables.ai and conflict checking by Salesforce Agentforce, and no subprocessor list ties those parties to what customer data they see.

Commercial Transparency

Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.

There is no pricing page on litify.com and no pricing entry in its navigation, but the vendor's own Salesforce AppExchange storefront publishes a figure and a unit: a default plan at $200 USD per user per month, marked as the lowest starting price, with discounts noted for nonprofits. The listing also states that the package requires Platform Cloud, so a buyer can see that a Salesforce licence sits on top of the Litify subscription even though the size of that addition is not given. What is withheld is the rest of the range: no tier names, no feature splits between packages, no statement of what the AI capabilities cost, and nothing on implementation, which is substantial for a Salesforce-native deployment. A buyer can therefore learn the floor and the unit of charge without entering a sales process, and nothing above the floor.

Firm and Practice Coverage

Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.

Coverage is described with real substance. Four solution lines each have their own page and named buyer: personal injury firms handling high-volume intake, firms doing high-volume defence work for insurance carriers, consumer and business immigration practices, and corporate legal departments managing internal matters and outside counsel. The AppExchange listing adds government agencies and nonprofits as named segments, and the customer roster bears the breadth out, running from Brooklyn Defenders to the United States Postal Service. The claim that the platform works across fifteen-plus practice areas is not itemised, and no statement of where the product stops was located: nothing on firm size floors or ceilings, on practice areas that are not supported, or on jurisdictions outside the United States, though the platform-of-action framing and the Salesforce licence requirement both point away from the very small firm.