Contract drafting in the age of AI

The theme of the drafting of contracts is undergoing a profound transformation, driven by the advance of artificial intelligence. Indeed, a recent study by LegalBenchmarks showed a surprising result: generalist AI models, such as Google Gemini 2.5 Pro, outperformed legal tools that specialize in contract drafting, achieving a reliability rate of 73.3% versus 70% for the best human professionals. A figure that opens crucial questions: how is it possible for models not designed for the legal field to outperform solutions born for that purpose?

‍

The article analyzes the results of the study in detail, highlighting advantages and limitations of both technologies and focusing on the growing skepticism of part of the legal community, which fears an over-reliance on automated tools. At the same time, it explores widespread unease among young lawyers, who are increasingly concerned that AI will erode key skills such as contract drafting.

In this changing scenario, the importance of relying on qualified IT consultants and technology partners emerges, such as Lanpartners, capable of accompanying law firms and professionals in digital transformation, ensuring a balance between intelligent automation, data security and human control.

‍

‍

La drafting of contracts has always been the beating heart of the legal profession, an activity that requires precision, regulatory knowledge, interpretive sensitivity and, above all, the ability to mediate between opposing interests. Today, however, this balance is facing a profound transformation. Artificial intelligence has now become much more than simply assisting in researching or analyzing documents; it is beginning to draft contracts autonomously, with a degree of reliability that, in some cases, even exceeds that of humans. 

In fact, a recent benchmarking study revealed that generalist AI models designed for cross-use succeed in producing more consistent and accurate draft contracts than some specialized legal tools. According to data collected by LegalBenchmarks.ai, Google's Gemini 2.5 Pro achieved a reliability rate of 73.3%, compared with 70% for the best human lawyer participating in the test. 

The result, needless to hide, surprised the study authors and opens crucial questions: what does it mean for the future of the legal profession when a generalist AI model can outperform domain experts in contract drafting? What risks and opportunities lie behind this apparent technological “victory”? And how can a company or law firm consciously integrate such tools into its operational flow?

In the new frontier that now deeply unites advanced technology and law, the role of technology partners such as Lanpartners takes on an even more decisive value: accompanying studies, professionals and SME In the selection and integration of the most suitable digital solutions, ensuring security, control and business continuity. Because the future of contract drafting will not just be about AI, but about the balance between innovation and expertise.

LegalBenchmarks“ study and the ”surprise" result”

The study of LegalBenchmarks.ai, published in September 2025, represents one of the most in-depth comparative tests ever conducted on contract drafting using artificial intelligence. The research team analyzed thirteen different systems - Between generalist AI models, vertical legal platforms, and human professionals - subjecting them to thirty contract drafting tasks drawn from real use cases. In total, more than 450 outputs were produced and evaluated according to criteria of legal accuracy, terminological consistency, clarity and adherence to instructions.

What emerged overturned many entrenched beliefs in the legal industry. Indeed, the results show that human lawyers produced reliable drafts, on average, in 56.7% of cases, while some AI models exceeded this threshold. The model Gemini 2.5 Pro, for example, obtained a reliability index of 73.3%, slightly higher than the best human lawyer in the test (70%). Even more interesting is the comparison with vertical legal tools: platforms such as August e Brackets, designed specifically for contract drafting, stopped at 66% and 64%, respectively, demonstrating lower performance than generalist IAs.

The one parameter in which specialized legal instruments excelled was the workflow integration, that is, the ability to operate directly within software such as Microsoft Word or in document management systems. However, the quality of the content generated was less robust than the generalist models. In essence, vertical tools offer better compatibility with the day-to-day processes of a law firm, but not necessarily better quality of the drafting of contracts in terms of accuracy, consistency and completeness.

This figure, which may appear marginal, is actually the focus of the debate: if generalist AI produces more accurate drafts than those prepared by specialized legal tools or even experienced professionals, then the lawyer's role in contract drafting needs to be rethought: not eliminated, but evolved toward a supervisory function

Why generalist AI models have won in contract drafting

The surprising result-that models not specifically designed for the legal environment outperformed vertical models in contract drafting-requires some explanation. 

First, generalist models often have very large and up-to-date datasets, which include natural language, various formulations and numerous examples of contract documents. This very extensive training enables them to quickly “understand” contract structure and terminology.

Secondly, the development curve of the Vertical AIs can be slower: the specialized legal tools require training on specific regulatory and contractual data, which are often not updated as quickly as generalist models. This can result in a delay in “knowing” new wording or emerging clauses.

Third, in drafting contracts with standardized tasks, e.g., drafting of NDAs, supply contracts, SOW (Statement of Work), variability is less and automation is easier: In these contexts, “simple” AIs can perform very well.

For a legal department involved in high-volume contract drafting, these capabilities translate into saved time, increased consistency, and potentially reduced costs. However, it is crucial not to confuse the “draft” with the final product: legal contract drafting still requires a level of supervision and judgment that only the professional can guarantee.

Skepticism and limits of AI “victory” in contract drafting

While brilliant in numbers, the benchmark results require critical analysis and proper context: this is not a total triumph, nor is it an automatic replacement of human expertise in contract drafting. First, the study itself admits of methodological limitations: The choice of tools tested included some niche tools (August, Brackets) and may not represent the entire landscape of the legal technologies specialties available. 

Moreover, drafting a contract is not just about generating a first draft; it also requires interpretation of context, negotiation, management of the parties and their needs, and business assessments. In these areas, humans still retain a considerable advantage: In fact, the study finds that professionals excel at complex tasks such as integrating multiple inputs (templates, emails, communications), operations that AIs do not always handle well. 

Added to this is the issue of accountability: if the AI-generated draft contains errors, ambiguities, or fails to capture a hidden risk, legal responsibility remains with the adopting lawyer or legal department. For contract drafting, this means that AI cannot be “overridden” by human oversight, and clear protocols for review and control must be established.

Lawyers' fear and changing role in contract drafting

The digital transformation in contract drafting is generating some anxiety among young legal professionals: for example, About 34% of lawyers under 35 cite technology as the main threat to their careers. Why this fear? Here are some of the main elements that agitate the profession:

• La drafting of contracts represented a fundamental training skill for young lawyers: writing drafts, dealing with clauses, negotiating were key stages of professional growth.
• With AI that can generate reliable drafts, the risk is that that “on-the-job training” will be reduced or altered, changing the training path.
• New skills need to be developed: understanding of tools, oversight of IA outcome, ability to define effective prompts, quality control and integration of digital processes.
• There is a fear that AI-generated output will be taken as a “valid draft” without sufficient scrutiny, resulting in reputational risks as also explained in this article of ours.
• Law firms and corporate offices that do not offer adequate tools (hardware, software, training) expose young professionals to a technology gap, As has been explained here.

These concerns are closely intertwined with the theme of the drafting of contracts In the digital age: Indeed, the role of the lawyer is increasingly shifting toward coordination, supervision, and analysis rather than manual generation of the contract text.

What strategies for law firms and corporate offices in technological contract drafting 

To effectively address the change that is taking place in contract drafting, the law firms and corporate offices must adopt conscious and structured strategies. Here's how:

1. Evaluation of the instrument: select not only based on the quality of the draft (accuracy, consistency, legal sufficiency) but also considering integration with existing workflows (e.g., Word, clause management, repository).
2. Definition of human supervision: clearly establish the required level of control over IA output, such as review, modification, approval by the qualified professional.
3. Update of contract libraries: even with AI, it is vital to keep clauses, templates, and contract precedents up to date, because contract drafting remains an activity that invokes substantive and contextual knowledge.
4. KPI measurement: e.g., average drafting time, percentage of drafts approved without editing, number of errors found, satisfaction level of reviewers.
5. Training and change management: prepare professionals to use digital tools, understand their limitations, know how to interact with AIs, and handle critical reviews.
6. Technology-legal collaboration: Involve the IT team and legal department from the beginning to define flows, roles, responsibilities, and level of automation.

In the context of contract drafting, the adoption of technology should be seen as an evolution of process: thanks to AI, practitioners can focus on higher-value-added tasks, letting more standardized steps be expedited. This is where the following comes in. a high-level consultant such as Lanpartners, which can assist in technology selection, process design and team training by combining, through our experience, technological expertise and legal compliance.

Contract drafting: toward a balance between technology and human expertise

La drafting of contracts of all kinds remains a core competence for law firms and corporate legal departments, but the nature of that competence is changing profoundly. 

The advent of generalist AIs has thrown up a challenge and an opportunity at the same time: digital tools do indeed allow drafts to be generated with increasing efficiency and quality, but they require constant supervision, thorough checks and integration with the legal and organizational environment. The fears of junior lawyers are understandable, but the real benefit will be for those who can turn digitization into an opportunity for development, not a threat.

Therefore, addressing those seeking support in the digital transformation of contract drafting, Lanpartners offers an integrated approach ranging from tool evaluation to workflow definition, from team training to results monitoring. In an evolving world, it is those who can lead the change-and not undergo it-that will make the difference.