GenAI and the Future of Contracts as Strategy

A contract serves legal and operational purposes: it allocates risk, records an agreement and helps people coordinate work. Generative AI may assist with that communication, but the benefit must be demonstrated. Faster drafting is not evidence of fewer disputes, better collaboration or compliance.

Corrales Compagnucci, Haapio and Fenwick's 2025 introductory chapter “Generative AI and the Future of Contracts, Law and Design”, which opens the book they edited, Generative AI, Contracts, Law and Design, brings together proactive law, human-centred design and responsible AI. It is a conceptual framing of the field, not a controlled evaluation of AI-assisted contracting across organisations.

Design for the decision and the person making it

Identify who must use the contract: legal reviewers, procurement staff, operational managers or finance teams. Ask which action each person needs to take and what information determines it. A delivery team may need a trigger, deadline and notice recipient, alongside the clause governing exceptions.

Passera and colleagues' contract-diagram research reports comprehension benefits in an experiment with contract professionals. That supports testing information design. It does not establish that AI-generated diagrams preserve legal meaning or improve every business relationship.

Use headings, role-based guides and visuals where they solve an observed problem. Retain the source agreement and explain the relationship between a guide and the binding terms. Test readers' answers rather than assuming an attractive document is understood.

Use proactive design as an aim

Proactive contracting asks how an agreement can support intended outcomes and prevent foreseeable difficulties. Treat those as objectives to evaluate. An AI system needs relevant evidence before its predictions about disputes, delays or market practice can be trusted.

Cao and Lumineau's meta-analysis of contractual and relational governance suggests that formal contracts and relationship mechanisms can complement each other, depending on context. It does not show that an AI-generated contract creates trust. Include the people, incentives and working arrangements around the document in the assessment.

Give each AI use a separate evidence requirement

A suggested amendment is not an agreed amendment. Keep decision authority and any external action under the process appropriate to the mandate. A monitoring alert should prompt a decision by the person with authority; it should never change a commitment by itself.

Classify the actual AI use

The EU AI Act does not classify every litigation prediction or compliance-advice tool as high-risk. Assess the scope under Article 2, then the classification rules in Article 6 with the intended uses listed in Annex III, and the provider or deployer role. A tool for ordinary private legal research differs from a system intended to be used by a judicial authority or on its behalf, or in a similar way in alternative dispute resolution, under Annex III, point 8(a).

Serving an EU client alone does not answer every scope question. Data protection and professional secrecy also require their own assessments. Use the classification guide to connect duties to the actual use and applicable dates.

Run a redesign pilot with measurable outcomes

Choose an approved contract and a specific comprehension problem. Compare the existing version with a reviewed alternative in the relevant languages. Check correct task completion, overlooked exceptions, time and revision effort. Preserve legal review and record disagreements between users or reviewers.