Recruit
Artificial intelligence can identify a deleted word in a fifty-page agreement within seconds. But can it determine whether that word is worth delaying a strategic transaction or damaging a valuable client relationship?
This is no longer a hypothetical question. According to the Thomson Reuters Institute’s 2026 AI in Professional Services Report, 47% of the corporate legal departments surveyed are already using generative AI. This level of adoption is consistent with how legal professionals are using the technology: primarily for legal research, document review, and document summarization, repeatable tasks in which AI can save time and streamline working processes. Yet adoption alone does not answer the more important question: what should we delegate to technology, and what must remain human?
The case for AI begins with speed. It can process large volumes of information, summarize lengthy agreements, and compare multiple drafts, including versions that have been amended without tracked changes. During a negotiation involving several rounds of revisions, this can help prevent a small but significant amendment from going unnoticed.
Its contribution goes beyond document comparison. AI can help improve the clarity and coherence of contractual language, identify ambiguities, and detect inconsistencies involving defined terms, dates, party names, and cross-references. When used with approved templates or contractual playbooks, it can also flag departures from standard positions and prepare an initial list of issues for legal review.
This may create the impression that AI can perform almost the entire review. But it cannot independently determine what matters most to the business.
A commercial contract is not simply a collection of clauses or an allocation of risk. At its core, it is the framework for a relationship between people who have decided to build something together. Legal entities may sign the document, but behind them are individuals with expectations, responsibilities, commercial pressures, and a shared interest in making the relationship work.
This human dimension matters. A legally unfavorable clause may nevertheless be commercially acceptable depending on the value of the transaction, the importance of the client, the services involved, and the organization’s risk appetite. Conversely, wording that appears harmless in isolation may create a serious operational problem when considered in context. The best legal outcome is not always the strictest contractual position; sometimes, it is the solution that protects the business while preserving trust and enabling cooperation.
These decisions require more than pattern recognition. They require negotiation skills, proportionality, empathy, and the practical judgment to know when to insist, when to offer an alternative, and when accepting a calculated risk is the right commercial decision. They also require an understanding of how people communicate, where their concerns come from, and what they may need from the relationship in the future.
A well-drafted contract must also anticipate difficult scenarios. Agreements are rarely tested when everything is going well. Their real value becomes apparent when a payment is delayed, a service fails, confidential information is disclosed, or the parties wish to end their relationship. AI-generated drafting may appear polished while remaining generic or disconnected from the realities of the transaction. Anticipating how a long-term relationship may evolve and creating protections that are both effective and workable still requires experience and human reasoning.
Accuracy presents another limitation. Generative AI can confidently produce nonexistent legislation, rely on outdated rules, apply the wrong jurisdiction, or misinterpret a provision. Even apparently straightforward outputs, such as clause numbers, defined terms, and internal references, must be verified. The consequences are not merely theoretical. A Brazilian court has already addressed a case involving fictitious AI-generated precedents, demonstrating the consequences of treating a plausible answer as reliable. As the American Bar Association has emphasized, responsibility for the work remains with the lawyer, regardless of the technology used to produce or review it.
Confidentiality is equally important. Commercial agreements may contain personal data, pricing structures, and sensitive business information. Before uploading a contract, one must understand how the relevant AI tool stores, processes, and protects data, including whether information is retained, used to train models, or shared with third parties.
AI can make the first review faster and more organized. It still can’t tell you what a client will actually accept, or what’s worth pushing back on. The lawyers who get the most out of AI aren’t the ones who resist it, and they aren’t the ones who let it drive, they’re the ones who use the time it frees up for the parts of the job that still depend on strategy, negotiation, and the human relationships at the heart of every deal.
References
American Bar Association. Formal Opinion 512: Generative Artificial Intelligence Tools. 29 July 2024. https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-opinion-512.pdf
Tribunal de Justiça de Santa Catarina. TJSC multa autor de recurso por jurisprudência falsa gerada por inteligência artificial. 18 February 2025. https://www.tjsc.jus.br/web/imprensa/noticias/-/asset_publisher/GP1QtxFaSsX0/content/id/21350888
Thomson Reuters Institute. 2026 AI in Professional Services Report. 2026. https://www.thomsonreuters.com/content/dam/ewp-m/documents/thomsonreuters/en/pdf/reports/2026-ai-in-professional-services-report.pdf
