M

Meet Gabbi

The models are the product, and every other part of Meet Gabbi exists to hold what they produce. The voice agent answers the firm's line around the clock, speaks to callers through speech to speech technology with a proprietary layer, GabbiEmotion+, that reads emotional cues and adjusts tone and pacing, qualifies the matter and transfers the call when the firm's rules say to. The same agent runs text and webchat conversations, checks the firm's calendar and books consultations.

Under a chase protocol it calls, texts or emails a new lead within seconds of the inquiry. A newer interactive video avatar puts a face on the agent to walk website visitors through forms. The company describes the whole as a system meant to replicate the duties of an intake employee. The lead management system and transcription dashboard categorize and display what the agent captured, and without the agent they would sit empty.

Meet Gabbi does not cite law. What matters instead is whether the record of a call is faithful and whether the agent's judgments about the caller are reliable, since both decide which leads a firm pursues. Accuracy is claimed and never measured. The transcription page claims industry leading accuracy and the voice page promises stellar transcription accuracy, with no word error rate, test set or method behind either.

The lead management system sorts every call as a new inquiry, co counsel referral, existing client follow up, court notification or vendor call. The agent qualifies cases from what callers say. Nothing describes how often those categories are wrong or how a firm would notice. GabbiEmotion+ is presented as detecting emotional cues in real time, with no account of how that detection is validated. The Security and Trust page sets one boundary: the AI transcript ends when a call is transferred to the firm, so the record covers only the agent's part of the conversation.

The agent is sold to act on the firm's behalf, and the limits on what it may do are not written down. The chase protocol has it call, text or email new leads within seconds. The text and webchat agent qualifies leads and books consultations directly into the firm's calendar, and the avatar guides clients through intake forms and answers their questions. The company describes the whole as replicating what an intake employee does.

There are real review surfaces. Every conversation is transcribed and available in the firm's dashboard, leads are categorized and visible in real time, and the text page says complex situations or those needing a human are transferred to the firm's team with full context. What is missing is the rule set. Nothing says what triggers a transfer, what the agent may not tell a caller, whether it can quote fees or discuss a case's merits, or whether a lawyer reviews bookings before they stand.

The Security and Trust page frames security as a shared responsibility, leaving the firm to set permissions and configure the agent.

There are named voices and a scale claim, but no result anyone measured. A press release from September 2025 quotes a Pittsburgh criminal defense attorney and a lawyer at Atlas Law by name. Both describe the benefit they expect rather than one they recorded. The integrations page says its direct connections are battle tested by hundreds of law firms, without naming the firms or giving a count. No case study, figure, before and after comparison or dated outcome was found on the home, voice, text, lead management, transcription, integration or about pages. No answer rates, time to first contact or signed case figures are published.

Specific commitments on separation, training and ownership, in a policy that never mentions privilege. The Privacy Policy (effective 30 April 2026), which section 9 of the Terms names as governing data handling, says each client operates in a segregated data environment with no access across clients, that client data is not used to train shared or global models or to improve models for other customers, that AI output is generated within client specific contexts, and that protected health information is not sent to external language model providers.

It leaves the firm owning its communications, transcripts, recordings, logs and AI output, with Meet Gabbi acting only as a processor, and offers business associate agreements where required. Data is shared only with service providers, with explicit consent or when legally required, and on that last route nothing commits Meet Gabbi to tell the firm before it discloses. Attorney client privilege and work product are not mentioned anywhere, and no retention period is stated.

The only statement on advice hands it to the firm, while the product is marketed on sounding human. Section 8 of the Terms (last updated 27 February 2026) says customers remain responsible for legal advice given to their clients and for compliance with legal advertising and privacy laws. Nothing says the agent itself gives no legal advice, although it qualifies cases, answers existing clients asking for case updates, and through the avatar helps clients understand legal processes and complete intake paperwork.

Nothing says callers, texters or website visitors are told they are dealing with AI. The voice page instead promises human like warmth on every call, and the about page describes an agent built to replicate an intake employee. The chase protocol and outbound campaigns described in section 6 of the Terms contact people by call and text within seconds, and nothing addresses consent for automated calls or the attorney advertising rules that govern solicitation, beyond placing those duties on the firm.

The Security and Trust page has a section headed Responsible AI and Data Handling, and it is about suppliers. It says Meet Gabbi evaluates the security and data handling of the providers in its platform and configures infrastructure with legal confidentiality in mind. It also declines to describe the platform as unhackable. It says nothing about how the agent behaves. No one is named as accountable for what the agent tells callers, and nothing describes testing before a change to its voice or qualification logic.

No finding is published on how GabbiEmotion+ reads distress, anger or urgency across accents, languages, ages or disabilities.

The Security and Trust page describes the controls in plain and modest terms, and the commitments that would hold the company to them are missing. It says the AWS hosted database uses AES 256 encryption at rest and data in transit is protected with TLS. Each firm's information stays in that firm's environment behind access controls, transcripts remain available until deleted under the firm's configured controls, and the AI transcript ends when a call is transferred.

The Privacy Policy names Supabase for storage and Twilio for voice and SMS as providers under business associate agreements, lists role based access and audit logging, and says all subprocessors are bound to confidentiality. A firm cannot find the rest. There is no full subprocessor list, since the language model and speech providers go unnamed. There is no retention period beyond as long as necessary, no breach notification commitment or timeline, and no deletion commitment at the end of a subscription. The page closes by calling security a shared responsibility and leaving retention requirements with the firm.

A two sentence limitation clause is the whole of the recourse terms. Section 12 of the Terms excludes lost profits, lost opportunities and indirect damages, and caps total liability at what the customer paid in the previous three months. For an intake agent, the harm a firm is most likely to suffer is a lost opportunity: a lead the agent mishandled, failed to transfer or never called back. That is the harm the clause excludes.

Section 6 adds that marketing outcomes are not guaranteed, section 11 disclaims any guarantee of uninterrupted service, and section 14 lets the terms change with continued use counted as acceptance. Nothing addresses what the agent says to a caller, a transcript that misrecords a matter, or automated outreach that breaches calling rules. No indemnity, service level, credit or warranty runs to the firm. The allocation is published and short enough to read in a minute.

Named connections with a line on what each one moves, and no documentation behind them. The integrations page lists native connections to CasePeer, Clio (two way sync with both Clio Grow and Clio Manage), Lawmatics (automatic lead creation and pipeline management), Lead Docket (real time lead sync including conversation history), Lawcus, MyCase, Salesforce and Zoho, and says the direct integrations are battle tested by hundreds of firms.

A custom integration team is offered for other legal software, and section 5 of the Terms ties custom integrations, intake workflows and chase protocols to a six month minimum commitment. The text agent also reads and books into the firm's calendar. The Security and Trust page says integrations are limited to the information the workflow needs and that connected systems' own practices govern data once it lands there.

No page documents the fields that move, the direction for most connections, the permissions the connection holds, or what happens when a record changes on the other side.

The tenancy model is stated plainly and the geography is not stated at all. The Privacy Policy describes a walled off architecture in which each client operates in a segregated data environment with no access across clients. The Security and Trust page says each firm's information is kept in that firm's environment rather than shared across accounts. The infrastructure named is an AWS hosted database, with Supabase for storage and Twilio for voice and SMS.

Nothing says which country or region stores recordings and transcripts, where calls are processed by the speech and language models, or whether any of it leaves the United States. No regional choice, private deployment or dedicated instance is offered. The customers are US law firms, and their calls carry health and injury details.

Alignment is claimed and an attestation is not. The Privacy Policy says Meet Gabbi maintains internal controls aligned with the SOC 2 Trust Services Criteria, and that formal certification status and documentation are available on request where applicable. It says the platform runs on HIPAA eligible infrastructure with providers under business associate agreements. HIPAA eligibility belongs to Supabase, Twilio and AWS rather than to Meet Gabbi.

No SOC 2 report, ISO certificate, auditor, audit period or penetration test is named, and there is no trust center. The Security and Trust page describes controls and says plainly that no provider can make security guarantees, but it cites no independent review. A firm can ask for documentation.

The infrastructure suppliers are named and the model suppliers are not. The Privacy Policy names Supabase for data storage and Twilio for voice and SMS, and the Security and Trust page names AWS for the database. A firm knows where its records sit and which carrier handles its calls. The models that listen, speak and decide are another matter. The Privacy Policy refers to external large language model providers and says they receive no protected health information.

The voice page describes speech to speech technology, and GabbiEmotion+ is presented as a proprietary layer on top. No language model, speech model, provider or inference location is named anywhere. The Security and Trust page says AI systems combine telecommunications, cloud, language models and integrations and that firms should understand how their information moves through them, and then does not say which models are in the chain. No notice of a change of model or provider is offered.

The contract mechanics are published and the price is not. The Terms describe subscriptions billed monthly, quarterly or annually, a 14 day free trial that converts to a paid plan unless canceled, and 30 days' notice to cancel. A six month minimum commitment with no early cancellation applies whenever custom integrations, intake workflows, chase protocols or outbound systems are built, which section 5 lists as examples of the work that triggers it.

Section 6 says advertising costs for marketing services are billed separately from service fees. A firm reading the Terms can learn how it will be committed before it learns what it will pay. No plan name, figure, rate per minute or per call, lead volume tier or setup fee appears on any page read, and the path to a price runs through a demo booking.

Meet Gabbi is built for law firms only, with a long list of practices and no stated edges. It says it was built specifically for law firms and names personal injury, criminal defense, family, estate, immigration, Social Security disability, employment, bankruptcy, corporate, real estate and intellectual property practices. The voice page describes the callers it is designed for, from potential clients after hours to existing clients seeking case updates and frustrated clients who need careful handling.

The integrations reach small firm systems such as Lawcus and MyCase and the intake platforms larger plaintiff firms run, such as Lead Docket and Lawmatics. The two named lawyers in its press come from criminal defense and a general practice. Nothing says what it does not do. No firm size range, call volume ceiling or language support beyond English is stated, and no practice area is named where the company advises against an automated agent, such as intake for a matter with a looming deadline. The breadth is asserted rather than shown.