T

Tonkean LegalWorks

The models drive a core capability on a product that plainly works without them. The AI is real and named: AI Front Door reads plain language requests, LegalGPT sorts and classifies unstructured requests, and packaged agents handle NDA generation and contract work. Underneath is a process orchestration platform built without code, with forms, rules, approval chains, dashboards and integrations, sold in the same form to procurement and IT.

The clearest evidence is contractual. The published subprocessor table lists Microsoft Azure for AI LLM engine services, OpenAI for AI and machine learning features and Google Cloud for OCR, and marks all three opt in. A customer that does not opt in gets an orchestration platform with no model behind it.

Grounding is described in outline and accuracy is disclaimed rather than measured. The retrieval story is stated: the AI Front Door trawls the data sources the customer has given it access to and generates a document if one does not already exist, and agents draw on connected enterprise systems, so a reader can tell in principle what an output is built from. Nothing beyond that exists. No accuracy figure, no test set, no evaluation, no error rate and no method description appears on any surface, and no output is presented with a traceable citation to the source record it came from.

The agreement moves in the opposite direction and does so explicitly: Tonkean does not warrant that the Platform will meet the customer's requirements or expectations, including with respect to any actions or outcomes of use of Tonkean's A.I. Bot. Nothing addresses hallucination in either direction. Searched the LegalWorks page, the security page, the pricing page, the DPA and the customer terms; docs.tonkean.com and the AI handbook were not opened.

The split between what runs alone and what reaches a person is published, and the customer sets the threshold. Simple requests such as NDAs and statements of work are handled automatically, and urgency is identified on every incoming request. Complex matters, including conflict waivers, intellectual property questions and outside counsel engagements, are routed to the right person or practice group. Approval paths and risk tiers follow the department's policy, intake forms adapt to risk level and matter type, and every action and approval is logged.

Because the thresholds are set by the customer, a buyer cannot learn from the published material what the system does by default. No confidence signal, abstention state or behavior on a misclassified request is described.

There is substantial evidence for the platform, and little of it is legal. The LegalWorks page shows twenty enterprise logos, including Google, Lenovo, AbbVie, Workday, Cisco and Intuit, under the heading trusted by enterprises like; it is the same strip used across the site, not a LegalWorks customer list. The headline figures have no attribution or basis: 50 percent shorter cycle times, 99 percent of customers reporting higher adoption, 30 employee hours saved weekly, 7.7 billion steps automated a year.

Two of three attributed quotes come from procurement. The Head of Global Procurement at Semrush gives the only hard figure, a cycle time falling from 19 days to 10, and the other is from the Head of Procurement at Cockroach Labs. The legal quote, from Mary O'Carroll, described as former President of CLOC and Head of Legal Operations at Google, concerns Tonkean in general rather than LegalWorks. A customer showcase exists and was not opened.

The marketing addresses privilege directly, and the agreement excludes the material it would protect. The LegalWorks governance section offers an audit trail of all actions and approvals, presented as helping preserve attorney client privilege, plus role based access control, ethical walls, item level retention and legal holds. The customer terms, however, say User Content is not information acquired by Tonkean, which takes customer content out of the contractual confidentiality duty and refers it to the Security section.

There Tonkean says it cannot fully ensure or warrant the security and privacy of User Content or personal information. The confidentiality a buyer can enforce over matter material is therefore thinner than the product page suggests, and the agreement governs. Two provisions point the other way. Tonkean has no right in User Content beyond the minimum needed to run the Platform and may not use it for any other purpose, and the DPA bars Tonkean from deriving rights or benefits from Personal Information.

A boilerplate disclaimer sits in a general agreement while the product is marketed in advice-adjacent terms. The customer terms state that the Service provides ideas, suggestions, analyses and other data for informational purposes only and not as advice, and disclaim responsibility for any information provided by the Service. That is a real sentence but a generic one, drafted for a business management platform rather than for a legal product, and the agreement carrying it was last updated May 2019, before LegalWorks and every agent now sold.

Against it the marketing describes agents that handle complex contract analysis and negotiation strategies, and an AI Front Door that answers employees' legal questions directly. The audience limits the exposure and is stated plainly: the buyer is a corporate legal department and the requesters are its own employees, so there is no consumer surface and no unlicensed practice question of the ordinary kind. What is absent is everything above the disclaimer: no rule of professional conduct or bar guidance named, no jurisdiction limit, and nothing on how a legal team supervises an agent that answers a business question without a lawyer seeing it.

No governance position for the vendor's own models is published, although its agents act on simple requests without a lawyer involved. There is no responsible AI page, principles statement, named owner for model behavior or testing regime before release. Nothing addresses bias, including whether triage and classification perform evenly across request types, business units or languages. That matters because the product's core job is deciding which requests are simple enough to resolve without a lawyer.

The security page is thorough but covers information security, not model governance. An AI handbook page in the navigation, headed with questions about what an agent is and when one counts as an enterprise agent, was not opened and may say more.

Almost every element is published, and one is weak. The customer controls retention in detail, setting item level policies down to individual fields so data is kept only as long as a process needs. The DPA adds that within 60 days of termination Tonkean will delete or return all Personal Data at the customer's choice and delete existing copies, keeping one copy only where law requires. Access control is described as comprehensive and role based across data access and process creation, and encryption is AES-256 at rest and in transit.

Audit logs that cannot be repudiated record all edits and processed transactions. The subprocessor list is published, naming each provider, its function and its country. Incident practice is the weak point: Tonkean commits to notify without undue delay after becoming aware of a Data Incident, with no stated deadline. The same clause bars the customer from publishing anything identifying Tonkean about an incident without prior written approval unless legally compelled.

Liability is addressed only through a limitation clause, and that clause disclaims the AI's exposure by name. Total liability is capped at the fees actually paid in the three months before the claim arose. Excluded damages are broad, covering loss or corruption of data, lost profits and pure economic loss. No indemnity runs to the customer. The only intellectual property remedy is that Tonkean may, at its discretion, procure a license, modify the Platform or terminate and refund the remaining period, and the agreement says no other rights or remedies accrue.

The indemnity that exists runs from customer to Tonkean. The warranty is to repair or replace material errors that prevent ordinary use, and expressly does not extend to any actions or outcomes of use of Tonkean's A.I. Bot. No insurance position is published. The agreement was last updated in May 2019, so it predates the 2026 agentic product it governs and mentions AI once, to disclaim it.

Real integrations into legal systems are named individually, with direction described only in outline. The legal connections are Ironclad for contract lifecycle management, TeamConnect for matter management and SimpleLegal for legal spend, plus DocuSign and Adobe Sign for signatures and SharePoint for documents. The wider set includes Salesforce, ServiceNow, JIRA, Slack, Microsoft Teams, email, Coupa and SAP, and the product is designed as an orchestration layer across them rather than a system to migrate into.

Contracts sync from CLM to electronic signature, approvals run through email and chat, data is pulled from billing systems, and work is coordinated with outside counsel platforms. Nothing on the pages read says what a legal team must set up, which fields map, or what an integration requires. A public integration library and developer documentation exist but were not opened, and may say more.

The tenancy options are published in full, and the region choice is not. Three options are described: a multi tenant public cloud on AWS managed by Tonkean, a single tenant dedicated cloud on AWS also managed by Tonkean, and self hosting in the customer's own AWS, Azure or Google Cloud environment. The pricing page says multi tenant is included, while dedicated, customer cloud and on premise options are on the enterprise plan at extra cost.

Processing location can be worked out from the DPA. Every subprocessor is listed in the United States except Tonkean Israel Ltd., and Schedule 2 incorporates the 2021 Standard Contractual Clauses with UK and Swiss addenda for transfers out of the EEA. No region choice is published for the managed offering, and no page says where data is stored as distinct from processed.

Certifications are real, named and scoped, and the evidence is available only to customers. The security page states SOC 2 Type 2 compliance, independently audited under the AICPA Trust Services Criteria for security, availability and confidentiality. ISO/IEC 27001:2022 is stated as achieved, and a HIPAA compliance page sits alongside. No auditor is named and no audit period or report date is given for either. The reports are expressly limited to existing customers, who are invited to ask the team for the latest one, so a prospective buyer has no route to the evidence at all. There is no trust portal. The badges shown are the certification marks themselves.

The providers are named with their function and location, change notice is contractual, and the models themselves are not identified. Schedule 1 of the published DPA lists each subprocessor with its service and country: Microsoft Azure for AI LLM engine services, OpenAI, LLC for AI and machine learning features, and Google Cloud Platform for OCR, all in the United States and each marked opt in. It also lists AWS for cloud storage, Elastic.co for the hosted index database, SendGrid for email and Tonkean Israel Ltd. as an affiliate.

Tonkean must give notice before authorizing a new subprocessor, the customer has seven days to object, and an unresolved objection allows termination of the affected services. Azure and OpenAI are providers, not models, and no model or version is named, so a buyer cannot tell which model reads a legal request.

The unit and structure of charging are published carefully, and no figure appears. Charging is per Monthly Tracked User, defined precisely: a unique person who in a calendar month submits a form, opens a workspace app, submits or replies to an item in Slack or Teams, sends or replies to a Tonkean email, or clicks a button or updates an item. The vendor says these are not named seats or a count of records, that a counted person's actions that month are unlimited, and that billing uses the average across the subscription year so busy months even out.

A quote has three parts: the platform with unlimited workflows and connectors, the user volume, and optional deployment and services. Hosting and support tiers are itemized. No price, band, minimum or term is published, and every route ends in a scheduled call. A buyer can model the shape of the bill but not its size.

The buyer and the work are described in substance, and the edge of the product is left open. The segment is corporate and government in house legal departments at enterprise scale, with material addressed separately to employees raising requests, the legal team, and colleagues in procurement, sales and IT. Coverage is expressed as request types rather than practice groups: NDAs, statements of work, contract review, conflict waivers, intellectual property questions, outside counsel engagement, legal mailroom, email triage and general matter intake.

Customers are drawn from Fortune 500 and Fortune 200 enterprises. Nothing addresses use by law firms, no minimum department size is given, no jurisdiction or language coverage is published, and no request type is named as unsuitable for automated handling.