Stepping into the future

Why the interest?

Why have ChatGPT and other forms of generative AI received so much interest? First, ChatGPT is freely available for anyone to test out. By simply entering prompts or commands, it could seemingly address almost any question or execute any task that you ask of it. Users discovered that it could effortlessly help them perform various familiar tasks, such as writing prose or poetry, creating a resume, drafting an email, and even writing computer code. 

ChatGPT’s simple format has democratized the public’s access to machine-learning tools. It takes instructions and allows users to respond with further directions in plain conversational language. This capability made it easy for people to see its value, which was suddenly available to anyone — not just those who work with machine learning.

Generative AI has entered the collective legal consciousness quickly. In a recent survey conducted by the Thomson Reuters Institute, ChatGPT and generative AI awareness is significantly higher among legal professionals, with 91% of respondents saying they have heard of or read about these tools.

With that awareness comes the realization of potential use cases for using this technology in day-to-day work.

How does it work? 

Why do new forms of generative AI seem to be so effective at reproducing human intelligence? 

Machine-learning systems are not “thinking” systems. They are language prediction systems. Many machine-learning tools are designed to predict missing words in a sequence. When humans review those predictions for accuracy, they’re given a score for accuracy. The machine uses those scores to improve its predictions the next time. 

We have known for some time now that, as the data sets that machine-learning systems are trained on have become larger, their predictions and ability to provide accurate answers have improved only modestly. The big breakthrough with ChatGPT was that when the training data sets reached a certain size, there was an abruptly significant leap in accuracy. That increase in accuracy is the leap forward that will likely push generative AI into the mainstream. 

Generative AI in legal: Risks and opportunities

The legal profession is in a vital position relative to GenAI. First, legal work is primarily centered around words, documents, and data. The work consists of finding, analyzing, and creating texts and documents, which are precisely the tasks that generative AI is good at. 

Second, legal is a field where accuracy and precision are essential. There is little room for error when legal processes and decisions can impact parties' important rights or financial interests in legal agreements or disputes. 

Legal audiences are clearly impressed by the potential power of generative AI but also concerned by the dangers, as seen in the recent Thomson Reuters Institute report.

While ChatGPT can deliver answers that sound coherent and accurate, this type of AI is really just a sentence-completion engine. ChatGPT has little intelligence but is very good at drafting language that sounds like plausible responses to a user’s prompt. But it’s not always accurate! There are numerous examples of ChatGPT failing to answer questions correctly in fields — like law — where accuracy is essential. 

Those examples are called “hallucinations” and include the now-famous case of a New York lawyer who submitted a brief supporting a motion to a court with arguments and citations based on the results of a ChatGPT session. The brief included citations of non-existent cases and other legal inaccuracies, but it sounded plausible. 

Errors like that present an unacceptable level of risk for legal professionals. So, how will lawyers effectively use generative AI while maintaining all their client obligations? 

A standard methodology to improve generative AI results in specialized fields is called retrieval-augmented generation (RAG). In systems that use RAG, the user’s prompts or queries do not pass directly through to the LLM; the question runs first as a search against a trusted body of content — for example, verified legal content from a legal publisher or trusted documents from the user’s organization. Documents relevant to the question are retrieved first, and the question and the verified content are passed on to the LLM for processing. This process ensures that the answers to users' queries are grounded in trusted domain-specific data, not just a sample of data from the wider internet. 

The power of generative AI derives from the size and scale of the original data sets that the AI models are trained on. But, in specialized fields like law, it’s critical that training of the models takes place on trusted, verified, domain-specific data. 

What’s driving the adoption of AI in the legal profession?

A “perfect storm” of factors is pushing lawyers and legal organizations toward greater adoption of AI-based tools in their work. The main drivers of adoption include the following. 

Tasks centered on documents and data are likely candidates for automation

AI will likely have the least impact on labor in fields rooted in the physical world that can’t easily be digitized. Think in-person healthcare delivery, construction, and food service. Conversely, it has the most significant impact in fields that involve lots of documents and data and where the work outputs tend to be in the form of more documents and data, like banking and finance, media, design, and legal. 

Goldman Sachs has evaluated the potential impact of AI on jobs in numerous industries. Because it is such a data- and document-centric profession, legal is considered one of the industries in which AI will complement or replace a large share of work. Indeed, legal professionals will automate some tasks, but the fundamental transformation will be how their work is reimagined and improved. AI presents an enormous opportunity for legal because lawyers can enhance so much of the work they do with AI, thus improving client outcomes.

Lawyers themselves see the possibilities for AI-based automation of legal work

Lawyers are no longer being dragged into the digital age, kicking and screaming. ChatGPT, perhaps more than any of the new technologies that have swept over the legal services industry in the past decade, has caught the attention of lawyers, and most see generative AI as part of their future. 

The Thomson Reuters Institute conducted a significant survey of lawyer attitudes towards ChatGPT and generative AI within law firms. While lawyers are skeptical of the accuracy of generative AI and cautious about confidentiality and security concerns, ChatGPT seems to have opened the profession’s eyes to what might be possible. “The realm of the possible has sort of shifted in a monumental way,” one law firm representative said in the report. 

Of the survey respondents, 82% see that they can apply generative AI can be applied to legal work, and a smaller but significant share (51%) think it should be applied to legal work.

The widespread availability of ChatGPT — and the ability of lawyers to try it themselves and see the results in person — is undoubtedly behind much of this belief that generative AI could play a role in their work. When lawyers see a machine drafting a coherent document right before their eyes in response to a prompt they generated themselves, it’s hard to imagine AI not augmenting some of the tasks that human lawyers perform today. Indeed, concerns about accuracy and security will persist, but the hands-on nature of ChatGPT has convinced many lawyers of the likelihood that AI will play a role in their future work. 

Corporate legal departments are spending more, particularly on practice areas involving large sets of data

The prospect of a recession and general market uncertainty presents many risks to corporations, and their law departments expect those risks to lead to higher demand for legal services. The 2023 State of Corporate Law Departments (SCLD) report from Thomson Reuters reflects this data. This survey of law department leaders shows that the percentage of them planning to increase total legal spend is more than two times the number planning to decrease spending (41% vs. 20%). 

The focus of all this anticipated demand for legal services is not evenly distributed among practice areas. In particular, law department leaders see more demand coming from four practices: regulatory, labor and unemployment, disputes and litigation, and mergers and acquisitions (M&A).

What do those four practice areas have in common? Legal matters in those fields typically involve large amounts of data, including analysis of a company’s financial or operational data, large e-discovery projects, or large-scale due diligence reviews. So, it’s no surprise that law departments are turning to technology to carry an increasing share of the load in those areas. 

Using AI-based tools to augment lawyers’ work where large volumes of data are involved has already become mainstream in some of those practice areas. What’s new today is that awareness of new GenAI forms has inspired a higher level of interest in creative ways that lawyers can apply AI in legal practice to increase the efficiency and quality of their work. 

However, not all are responding to the challenge

Law firm Thompson Hines has tracked clients’ views on their outside law firms’ ability to innovate in their delivery of legal services. According to the 2023 edition of the Thompson Hines Innovation Survey, there is still a significant gap between the innovation clients expect from their law firms and what they are getting. The survey respondents include law department leaders and legal operations staff, who are often the personnel most in touch with leveraging modern technology in innovative ways. 

Where are clients looking for innovative technology from the law firms they work with? Results from the Thomson Hines survey largely align with the practice areas identified above as sources of increased legal spend. Clients seek innovations and efficiency in data-rich practice areas and legal tasks like e-discovery, document drafting, due diligence, transaction management, and legal research. 

Because they involve large data sets, these practice areas are precisely where various AI-based tools, including generative AI, can make the most difference for law firms. As the Thompson Hines data shows, firms are not yet meeting client expectations. 

According to the report, 96% of respondents wish their law firms would use innovation to save them money, yet only 12% say their primary outside law firms have done so. The most significant gap is in new technologies to improve processes or drive efficiency, which 72% of respondents want, but only 15% say law firms deliver.

Using data and technology to improve the quality and efficiency of the legal services delivered to clients is the most significant opportunity for law firms to leverage AI. Leveraging this technology has the most significant impact in areas where the capacity of humans to perform legal work has not kept pace with the explosion of data. 

Human expertise is critical to the success of AI in the legal field

It’s common to hear concerns that AI and other technologies will cost lawyers jobs or diminish the value of their work. Generally, while AI might entail a shift of some specific tasks from humans to machines, the idea that AI will lessen the contributions of human lawyers could not be further from the truth. The opposite is true: AI requires legal domain expertise and skills to function correctly. 

It’s common to think of AI as a technology. But it’s more helpful to think of AI as the convergence of three resources: the data on which AI operates, the technology that makes it work for specific applications, and the human expertise that directs the technology to practical applications and solutions while informing training and error analysis.

AI won’t work in a specialized field like legal unless all three of these components are present, and none of them alone is more important than the other two. Here’s why:

• Data or content is critical because it is often both the input and the output of the process. Data is necessary to train machine-learning algorithms. AI-based legal technology solutions are built on large volumes of current, accurate, comprehensive, and enhanced content. Data drives our legal system — the statutes, regulations, case law, and other legal and administrative opinions collectively represent the data that lawyers and judges must research, analyze, interpret, and reason over. Without trusted, lawyer-created content, no reliable machine-learning tools would be possible.  
• Technology drives AI solution design and development, typically requiring robust, complex solution architectures that can work at scale. AI scientists combine expertise in AI tools and technologies, analytical and problem-solving skills, and a solid understanding of the target domains.
• Subject-matter expertise is crucial as it ensures that AI tools solve the correct problems and capture the nuances of the domain in a way lawyers can understand and analyze. Experts inform the creation of AI training data, validate the performance of machine-learning algorithms, and play a critical role in error analysis. They help computer scientists and engineers understand the domain and the data attributes responsible for errors so that they can be corrected.

Experienced lawyers with a comprehensive knowledge of legal workflows and the needs of their clients are in an excellent position to leverage AI to expand their role and deliver value. Lawyers spend their careers learning the law and providing solutions to clients; AI can help them amplify that knowledge and experience if they are willing to see the opportunities. 

How modern lawyers are putting AI to work

All three elements — data, technology, and legal domain expertise — are assembled in various configurations in today’s legal AI solutions. Here are some examples of how AI supports lawyers in their work today. These are mainstream applications of AI today, not some long-off future state. 

Contract management and analysis

AI is increasingly expanding lawyers’ ability to organize, monitor, manage, and negotiate large volumes of contracts. A modern contract management system enables lawyers to draft better contracts faster and simplify the review and analysis process by identifying critical risks, opportunities, obligations, and rights in agreements.

Reviewing and analyzing legal documents and contracts is both time consuming and critically important. The stakes are high, and mistakes can be costly. Legal expertise and interpretive ability are essential to the task.

Lawyers must separate the noise from the key facts and clauses; find document variations from negotiating parties; analyze single, stand-alone agreements; and assess numerous contracts that are part of large, complex transactions, mergers, and settlements. Sometimes, passages written differently mean the same thing; other times, adding or deleting a single word dramatically alters meaning.

Automating due diligence and compliance review through AI and machine learning-powered solutions can save time, reduce costs, and avoid errors.

By reviewing and extracting information from large volumes of contracts and tracking similar information across documents, AI-driven tools relieve lawyers of unrelenting administrative work, including hours of cut-and-paste extraction of clauses. These time savings free them to spend more time assessing and mitigating risk, identifying opportunities, and delivering essential insights to their companies and clients. 

The 2023 Thomson Reuters Institute Dynamic Law Firms Report found that firms that focus on fostering agility, making strategic investments, adapting to changing market demands, and prioritizing client-centric approaches experienced 5.7% growth rates in profit per lawyer calculations over the last 10 years. In contrast, those firms that did not take similar actions averaged a 0.8% decline.    

Legal research

Another cornerstone of sound legal practice — research — is also being transformed by artificial intelligence. Modern legal research platforms, utilizing intelligent search and document analysis capabilities, can save time, improve results, and inform advice and litigation strategy.

AI-powered legal research solutions like intelligent document analysis can review a lawyer’s or opposing counsel’s brief, compare it with arguments from past cases, identify weaknesses, and suggest additional highly relevant citations. They also can warn litigators about citing cases that have been implicitly overruled or abrogated but have no direct citations flagging them as invalid.

Legal research platforms have gradually enhanced question-and-answer capabilities so that research has become less of a search-and-retrieve process and more of a dialog between lawyer and machine, with more opportunity for nuance and focus in search results. ChatGPT and other forms of generative AI are set to accelerate that development further.  

E-discovery

The first mainstream use of AI in law was in e-discovery, in which hundreds or thousands of pre-trial documents can be analyzed to identify relevant sources of evidence for litigation. Machine learning has enhanced e-discovery by creating the ability to model relevant data and train the machine to identify relevant information, learning from existing data and feedback from subject-matter specialists. It does this in a fraction of the time and generally much more accurately than manual review. 

Data-driven litigation analytics

Informed by court dockets — and sometimes case law and other legal information — some research tools can display statistical information by judge, lawyer, law firm, company, and case type. These identify factors that impact outcomes and help guide legal strategy. Because AI-powered technology can access more relevant data at fast speeds, it is better equipped to predict the outcome of proceedings, ensuring lawyers are empowered to provide the best possible strategic and tactical advice.

Billing and timekeeping

AI technology can simplify billing and timekeeping by identifying line items in clerical and administrative tasks, such as time entry, invoice submission, and review. It can also review entries, verify them for accuracy, and highlight duplicate entries. This technology is also powerful enough to analyze billing data and answer questions such as what is the likelihood a corporate client will pay for a particular piece of work or what type of expenses are routinely paid and which are commonly rejected. This capability can be a valuable operational, planning, and budgeting tool.

Generative AI: The new imperatives for law firms and law departments

How should generative AI change how lawyers think about applying AI in their work? Here are some implications and imperatives that GenAI will require legal organizations to address.

• Trust. The best relationships between lawyers and clients always revolve around trust and confidence. Because generative AI is built on data and content, the transparency and provenance of data sources have become even more critical. Lawyers need to be able to demonstrate that the outputs from the AI systems they use are run on authoritative content and to work with technology providers who understand this. Large language models built on massive, diverse data sets work well for many general tasks, but lawyers require answers and outputs grounded on verified authoritative data. Generative AI “hallucinations” are not an option for lawyers. 
• Security and confidentiality. Like other legal technology forms involving client data, using generative AI will entail certain risks. Organizations will need to have policies that govern the use of client data in deriving new work products. Processes and data governance policies will be required to ensure that only authorized users can access client data. Using copyrighted material to build the models behind LLMs such as ChatGPT is giving rise to emerging intellectual property issues. 
• Integration of tasks. One enhancement that generative AI is likely to bring to the legal field is the integration of tasks that once were considered separate parts of a workflow. From a document draft, lawyers will be able to launch into a research question without leaving their draft. Conversely, while researching a point of law, lawyers will be able to instruct their research tools to initiate a draft. The boundaries between research, analysis, and drafting will become more fluid.  
• Exploit legal domain expertise. The winners among lawyers will not be the ones who passively wait for generative AI tools to come their way. Successful lawyers will understand that their domain knowledge is a competitive advantage as they help deploy AI systems that embody and amplify that knowledge. Lawyers will find new roles and opportunities to blend their legal knowledge with process management and service design. Lawyers’ work product will continue to provide the authoritative data on which their organizations’ AI systems run. 
• Greater interactivity. Anyone who has used ChatGPT understands the new paradigm it presents. Generative AI allows lawyers to engage in a dialogue with data and technology in new ways. Research tasks will entail back-and-forth refinement of results as the lawyer and machine learn from each other. Drafting will be faster and allow lawyers to quickly adjust and get the machine’s help changing tone, emphasis, or scope. 

Step into the future

In a wide range of applications, AI-powered tools are leveraging data, sophisticated analytics, and subject-matter expertise to transform the practice of law. Modern, forward-thinking law firms and corporate law departments use this technology to improve their results, unearth insights that minimize risk, and identify opportunities — doing it all with greater speed, accuracy, and efficiency.

Generative AI adds a new dimension to the successes already seen in applying technology to legal work. The machines are getting more powerful in the tasks they can accomplish, but the real story is the opportunity for lawyers to extend and amplify their legal expertise with the help of technology. 

Learn more about the future of AI at Thomson Reuters and how it can redefine your ways of working.