Skip To Content

Lawyers and the AI knowledge gap

Author(s): Simon Wormwell, Gillian S.G. Scott

Dec 13, 2021

While lawyers increasingly recognize the importance of legal technology, and specifically artificial intelligence (AI), for efficient and effective client service, there remains a significant gap between the anticipation of the impact of technology and an understanding of the technology itself. In the recent Future Ready Lawyer report, 70% of responding lawyers in corporate legal departments noted that AI will have an impact on their organization in the next three years. However, only 28% of respondents indicated that they understand AI technology very well. A similar gap in knowledge was found among responding lawyers at law firms.

In this article, we hope to reduce that knowledge gap by way of a practical review of some common legal work that can be supported by AI tools readily available today, much of which can be applied in contracting, a core component of any legal department.

What is AI?

AI can be described as the use of computers and software to replicate human decision making. This could range from the automation of simpler tasks to the exercise of human-like judgment. AI in legal practice today has recently been described as “better search and find” and “‘Control + F’ on steroids,” including when applied to the review of contracts. This highlights both the power of AI and its current limitations in the context of legal work.

How is AI commonly applied to contracts today?

The most common and effective applications of AI to contracts today are at the beginning and end of the contracting cycle. At the beginning, AI assists in the creation of first drafts (for example, through tools such as Contract Express, HotDocs, GhostDraft (Korbitec) and Leaflet). AI can be equally useful in the review of completed contracts (for example, through tools such as Kira Systems, eBrevia, Diligen and Luminance).

Document automation tools can be used to assist in the generation of first drafts of contracts. Beginning with appropriate template documents, various fields can be “coded” to prepare them for use – these effectively provide placeholders for users to apply common provisions in contracts that change. Once coded, the contracting tool receives values for each field from users, often through the completion of a pre-set form. Those values are then compiled into the coded template to complete a draft agreement. Though not often included as an example of AI, these tools replicate basic human decision making and automate related tasks.

The use of AI to support human review of completed contracts is now commonplace. The strength of these tools is in finding and categorizing requested types of clauses. A subject matter expert can then more easily review the identified clauses and exercise judgment. The time, cost and resources required to “manually” complete due diligence in transactions were the primary business drivers for the development of these AI-based contract review tools. There are now many commercially available tools that are pre-trained “out-of-the-box” to identify hundreds of common contract clause types, in many cases more quickly and accurately than human-only review. To varying degrees, these tools can also be trained by users to identify new types of clauses.

Common application of these tools is expanding beyond transactional due diligence. For example, large document sets can be reviewed to collate clauses for future contract drafting or to bring forward clauses for downstream contract management purposes. Data regarding contracts can be collected over time, providing risk profiles and informing future contract negotiations.

Developing use cases: contract review and negotiation

AI tools to support the pre-execution review and negotiation of contracts are becoming more readily available. AI tools can carry out a variety of baseline proofreading activities, reviewing cross-references, defined terms and definitional uses. Newer, more complex applications review clauses in a contract under negotiation to show variances from preferred forms of clauses. Optional language can be suggested for clauses varying from a standard. Automated comments or “redlines” of a document can be generated for user review.

In this way, AI tools are moving up the value chain to support more judgment-oriented contract work. From a practical perspective, tools for this phase of contracting initially focused on high volume commercial contracts (such as non-disclosure agreements). Increasingly, this technology can be applied to a variety of other contracts selected by the user. Whether AI tools are useful for a contract type may depend on whether there are enough suitable examples to enable the AI software to make valuable comparisons.

What are the benefits?

AI-supported contract assembly and review offer several potential benefits to users and their clients. The time required to draft and review contracts can be greatly reduced. This allows legal subject matter experts to focus on the delivery of higher value aspects of practice for their clients. AI contract generation can also reduce the overall cost of contract work. AI tools have been shown to complete tasks with greater accuracy relative to human-only review. Incorporating AI in contract review can reduce risk of error, thereby increasing client satisfaction. And an added benefit of having the support of technology to complete what are often the more mundane aspects of contract work is that it makes for happier practitioners.

Humans and AI robots

Hopefully, it is now trite to ask whether robots will replace lawyers, and equally trite to counter that they will not. Where a clear-cut decision-making process can be automated, it should be; this is not “lawyering” per se. However, most legal work in which clients see higher value requires that judgment be exercised, often in contexts not yet easily accounted for by AI. Common AI contract review tools can spot indemnity provisions in multiple locations in a contract, but judgment of a legal professional is required to assess the contextual risk to a client. The value proposition of AI in law is to enable legal professionals to more quickly and accurately complete certain complex high-volume work as well as common repeatable tasks. The measure of success for AI in legal is whether it enhances the delivery of client service by the legal subject matter experts themselves.

Streamlining legal work is effectively a process improvement project. Like any technology, AI is simply a tool to help improve a process. Before engaging with technology, users should ensure they have a clear understanding of existing processes and desired improvements. Automating a flawed process may amplify inefficiencies. If AI technology is the right tool, users can expect to invest significant upfront time and expertise in configuration, in addition to the cost of licensing the technology. Even the simplest use case for document automation technology requires considerable work to prepare templates for automation. Anticipated gains in efficiency must be measured against upfront investment.

Maximizing return on investment from AI technology might also mean specialist staffing. When AI contract tools are used at scale, larger legal organizations and alternative legal service providers may have dedicated personnel with legal and technical expertise to use or support the use of the technology. Each tool has its own user interface, functionality and workflows which can generate greater value if the technology is used by an expert. At the same time, it is important that any specialists dedicated to this work also be well-connected to or integrated with the client service teams they support; they cannot exist in silos.

For some use cases, users may also have to make a significant investment in “training” or “feeding” AI technology at the front end. Commercially available tools for the AI supported review of contracts generally come ready or “pre-programmed” to identify certain contract clauses. To apply such AI tools to new clauses may require extensive training, including loading significant sample volumes into the tool. This may also require legal subject matter expertise to validate or correct the findings of the AI tool as it learns.

The number of samples required will vary depending on the use case. In the example of automated generation of comments noted above, 100 or more suitable contracts might be required to establish a standard or “playbook” from which consistently valuable comparisons or redlines can be generated. That number may be reduced over time through enhancements in the technology.

Humans aren’t perfect, neither is software

Like humans, software systems are imperfect. AI tools considered “market ready” do not perform perfectly. This is acceptable, provided users understand their limitations. When lawyers hesitate to use AI, it is sometimes because they believe the outputs should be perfect; but this isn’t (and can’t be) a practicable objective. Legal professionals are beneficiaries of efficiency gains in their capacity as users of AI review tools, but they also play a quality assurance role.

Users should expect to work closely with legal technology companies to understand and refine applications of AI. Lawyers should also be transparent regarding the use of AI tools with their clients.

Looking ahead

Though the group of professionals in the legal industry who need to be functional experts in AI can be small, a critical mass of lawyers who are versant in the possibilities and limitations of AI in the legal space is necessary for AI to become more widely adopted (and for organizations to reap the benefits). While change management will be no mean feat for organizations in implementing AI technologies, scaling that mountain now will ensure that legal departments and their clients will improve their centralized data and be better prepared for the other uses of AI that may be coming.  

Additional information regarding a number of these and other legal technology tools can be found at the Legal Technology Hub.