AffiniAI (Beta)

AffiniAI is a purpose-built contract analysis assistant integrated into the Affinitext platform. It is designed for a different purpose from general-purpose AI tools.

While AffiniAI uses a large language model — currently OpenAI's GPT-4.1 — the model is only one component of the solution. AffiniAI combines the language capabilities of the model with Affinitext's proprietary workflow, including semantic search, structured document navigation, and verified cross-references and defined terms, to provide answers that are grounded in your actual contract text.

Unlike generic AI tools, AffiniAI does not require users to upload full contracts into a general chatbot workflow. Your complete contract documents never leave the secure Affinitext environment. To answer a query, AffiniAI sends only the minimum necessary excerpts, such as relevant clauses or defined terms, for processing by the model via OpenAI's API.

When you submit a query, AffiniAI follows a structured multi-step workflow — it does not simply pass your question to a language model and return whatever comes back.

i. First, it performs a semantic search across your Affinitext library to identify clauses relevant to your query.
ii. Next, it follows linked cross-references and defined terms using the structured relationships held within the Affinitext platform.
iii. It then assembles the relevant context needed to answer the question and sends only those targeted excerpts to the language model.
iv. If the model requires additional information, AffiniAI retrieves further content and repeats the process.
v. Finally, it presents the answer with clause-level traceability so you can review the supporting contract text directly to validate the answer.

Your full documents and libraries remain exclusively within the secure Affinitext environment at all times.

This is because AffiniAI is doing genuine analytical work, not just generating text from a prompt. Each query triggers a multi-step process: semantic search, clause retrieval, cross-reference and defined-term lookups, and iterative context assembly before a response is generated.

This is by design. AffiniAI prioritises accuracy, traceability, confidence and legal defensibility over raw speed. The result is an evidence-backed answer linked directly to your contract text, rather than a fast but potentially unreliable response.

AffiniAI currently uses OpenAI's GPT-4.1 as its LLM, which is hosted on OpenAI's infrastructure, generally in the United States.

Only the minimum, essential excerpts of contract text are securely and temporarily transmitted to generate a response. Your full contract documents stay within the Affinitext environment at all times and are not sent to the model.

Your full documents, contract libraries, and related files remain within the Affinitext environment and are not sent to the model provider.

To answer a query, AffiniAI may securely send only the minimum necessary excerpts, such as specific clauses or defined terms, to the language model for processing. These excerpts are used only to generate the response to your query. This data is used solely for that purpose and is not retained, stored, or reused by the LLM provider after the session ends.

No. Affinitext does not use your data for training purposes. All documents, queries, and interactions are processed solely to provide you with accurate responses and insights. Your data is not stored, reused, or shared for AI model training, ensuring your sensitive contract information remains confidential and secure.

AffiniAI stores each user's chat history within the secure Affinitext environment, so you can revisit and continue previous conversations. This chat history is visible only to you. You can rename or delete your chats at any time — once deleted, they are gone for good.

AffiniAI is designed to minimise data exposure during query processing. Rather than sending whole documents to the model, it sends only the minimum necessary excerpts required to answer the query.

Data is transmitted through encrypted channels, and access to AffiniAI is governed by the same permission controls that apply across the Affinitext platform.

AffiniAI is built on the Affinitext platform and applies the same underlying security and access controls as the rest of the platform, including:
• Data encryption in transit and at rest
• Role-based access controls• Document-level permissions
• Regular security audits and penetration testing

Only authorised users can access AffiniAI, and users can only see content they are already permitted to access within Affinitext.

Affinitext upholds and complies with vital security certifications, including ISO 27001:2022, Cyber Essentials, Cyber Essentials Plus, and UK Ministry of Defence Accreditation (OFFICIAL SENSITIVE).

OpenAI API and business offerings are covered by a SOC 2 Type 2 report and are certified to ISO/IEC 27001, 27017, 27018, and 27701:2022.

We chose GPT-4.1 because it offers advanced natural language understanding and processing capabilities, enabling AffiniAI to interpret complex queries and provide accurate, contextual responses — ideal for sophisticated contract-related questions.

However, the model is only one part of AffiniAI. The core value of AffiniAI lies in Affinitext's workflow around the model, including semantic search, structured navigation, verified relationships, and clause-level traceability. AffiniAI is designed to remain model-agnostic, so Affinitext can adopt other models in future where that improves performance, deployment options, or customer fit.

We welcome your feedback! Please contact our support team through Chat on the Affinitext platform or email us at support@affinitextgroup.com with any questions, concerns or suggestions.