C

Contractify

Ada reads uploaded agreements and related material, including general terms, addenda and nondisclosure agreements, to suggest parties, dates and renewal details. Users can ask questions about clauses or request a summary, then add the response to the shared contract record. These model functions turn unstructured text into information that drives tasks and reminders. The surrounding repository, approval routes, signature connections and access controls also perform ordinary contract management work.

Centralize includes an AI timeline; Collaborate adds summaries and contract chat; Control adds portfolio retrieval and custom prompts. Contractify therefore combines model assisted analysis with a broader workflow and storage application.

Suggested contract values point to the place Ada found them, allowing a reviewer to inspect the supporting text before saving the result. The assistant also draws on agreements and metadata in the customer's environment when answering questions about renewals, suppliers or individual clauses. These features ground extraction and contract questions in source material. They do not establish an independently measured error rate.

Version 15 of the service terms warns that automatically generated output can be incomplete or inaccurate and requires customer verification. A current reproducible evaluation of extraction accuracy, invented answers or performance across languages is not specified.

People verify and save Ada's proposed contract data, and automatic follow up tasks use the approved values. A user can place chat responses on the summary page and assign follow up to the responsible parties. Contract approval and signature flows add separate business review steps, while administrators control who can edit or manage agreements. The service terms require verification and human intervention, and prohibit using Ada for automated decisions under Article 22 of the GDPR.

This gives reviewers identifiable control points. A confidence threshold that stops a review, an AI failure escalation route and a correction procedure after an incorrect automated task are not specified.

Meat & More uses Contractify to centralize agreements and attachments, delegate tasks and receive notices for commercial leases and supplier contracts. Its legal case describes more than 200 agreements previously tracked in spreadsheets, over 50% less time spent on legal follow up and more than €40,000 avoided on sales agreements in one department. These are results for the contract management workflow, without a stated study date or measurement method.

Vivaldis also describes centralized access and timely notifications. Ada's separate efficiency illustration compares manual registration with assisted analysis, but does not identify a customer or a current model evaluation behind that comparison.

Customer information remains the customer's property, and the service terms impose continuing confidentiality obligations on both parties and their staff or subcontractors. Contract access can be restricted by individual agreement, folder or project. Section 3.4 permits model training only with prior, separate written consent, while the AI clause and DPA state a prohibition on using uploaded data for training. AI providers must also exclude training or keep it disabled.

Prompt processing can involve temporary retention for abuse prevention. Return and deletion obligations apply at exit, but attorney client privilege, work product handling and enforcement of source system ethical walls are not specified.

Ada's output is expressly described in the service terms as neither legal nor commercial advice, and customers remain responsible for verifying it. The same clause prohibits using the AI functions for automated decisions under Article 22 of the GDPR without human intervention. Product instructions give that responsibility a concrete step: a person verifies and saves proposed contract values before they drive follow up.

Legal teams can retain oversight while distributing routine tasks to other departments. Named professional conduct rules, jurisdiction specific supervision duties and limits on which legal questions the assistant may answer are not set out.

Contractify's July 2026 AI implementation guide describes ground rules for human review, permitted data and product limits, illustrated through Ada. It explains that a person validates extracted information and that answers draw on the customer's agreements and metadata. The account of Ada's initial pilot names time saved, extraction accuracy and correction volume as criteria. That historical pilot is separate from a continuing evaluation program for the current model stack.

An accountable AI governance owner, release approval framework and results on uneven output across languages or contract types are not specified. The security program concerns infrastructure and data protection.

AI subprocessors may keep prompt data for abuse prevention for at most 30 days, subject to longer legally required retention. The DPA requires return or secure deletion within 30 days after processing ceases or the agreement ends, with written certification. Routine database backups have several intervals, extending to an annual copy kept for one year. Access controls distinguish user roles and restrict contracts, folders and projects.

Named subprocessors have stated functions and locations, and new providers require notice. A suspected or known personal data breach must be reported immediately and within 48 hours, with investigation, mitigation and coordinated communications.

The terms promise material conformity with the agreement and give DB2 options to contest an intellectual property claim, modify the software or refund the unused fee after taking it back. General liability is capped at the greater of the previous 12 months' fees and €25,000, with exceptions for death, injury, fraud, intent and gross negligence. A confidentiality breach carries €5,000 compensation. The DPA separately provides mutual indemnities for its breach, limited by insurance payment or previous annual fees.

AI output still requires customer verification; a specific remedy for an incorrect AI answer and the insurer's policy scope are not stated.

JuriBlox and Clausebase provide drafting connections, while Nitro Sign and DocuSign support signature flows. The JuriBlox arrangement moves from template based contract preparation into Contractify approval, signing and follow up. The public API supports personal access tokens or OAuth authorization, and webhooks send resource identifiers that an integration uses to retrieve full records. Webhook signatures, retry timing and manual redelivery are specified.

These are usable components for connecting business systems. Detailed field mapping, permission behavior and failure handling for the named drafting and signature connectors are less explicit than the general API mechanics.

The hosted application keeps primary data in DigitalOcean's Frankfurt facility and backups in Amsterdam and Dublin. The current provider register distinguishes storage from AI processing: OpenAI is in the United States, Mistral handles text recognition and structure in France, and Microsoft Azure processes AI work in Switzerland. Other providers support signatures, customer service, analytics and communications. The DPA sets conditions for international transfers and incorporates named recipients.

The main storage locations are therefore more specific than a general European hosting claim, but customer region selection, deployment alternatives and residency differences between subscription packages are not specified.

Security controls include encryption in transit and at rest, database access restrictions, temporary file links, two factor authentication, activity logging and annual penetration testing. The security sheet names ISO 27001 and other certifications held by DigitalOcean, and points to Amazon's certificates. Those attestations concern hosting providers rather than a defined Contractify service scope. Contractify's terms offer current security material and any available independent reports on request once per contract year, while penetration test results are also available on request. A specific independent certification, audit period or assurance report covering Contractify itself is not identified.

The current subprocessor register identifies OpenAI and Microsoft Azure as generative AI service providers. It also names Mistral for optical character recognition and document structure analysis, and Meilisearch for vector retrieval. Processing locations distinguish OpenAI in the United States, Mistral in France and Azure in Switzerland. The terms require provider training to be excluded or disabled and limit temporary abuse monitoring retention.

Customers receive notice of new subprocessors under the DPA. Exact model families and versions, the routing of a particular task among providers and a commitment to notify customers of model substitutions are not specified.

Subscriptions charge for a package and seats, with annual advance billing. Centralize has a fixed maximum of five seats shared between users and viewers. Collaborate starts at five users and Control at ten, with user and viewer seats billed separately. Signing allowances are 200 and 500 respectively in those higher plans, while stored contracts are unlimited. Adding seats requires an administrator to confirm a billing warning; deleting or blocking a user does not itself reduce the contracted quantity.

Downgrades need 30 days notice before renewal. Subscription amounts, extra seat rates and implementation charges still require the customer's quote.

Legal, finance, procurement, management and IT teams use the service for their organization's agreements. Its legal workflow gives counsel oversight across entities while other departments own tasks and signing responsibilities. Named examples include commercial leases, supplier agreements, sales contracts and employment information. Ada handles contract data, summaries and questions, with language coverage described broadly rather than through a tested language list.

Draft creation also relies on connections such as JuriBlox and Clausebase. Specific law firm segments, jurisdiction boundaries and unsupported agreement types are not defined, and the standard license is for internal business use.