eDiscovery, legal research and legal memo creation - ready to be sent to your counterparty? Get it done in a heartbeat with AI. (Get started for free)
Law firms and legal departments are buried under mountains of documents. The discovery phase of litigation alone can produce millions of pages that must be reviewed for relevance and privilege. This labor-intensive process is a major cost driver that saps time from more strategic work. However, AI is providing welcome relief by automating document review tasks.
Rather than relying solely on armies of junior associates and contract attorneys, law firms can use AI to cull irrelevant and duplicated documents from review sets. Machine learning algorithms can be trained to identify concepts and fact patterns within documents. This enables technology-assisted review solutions to rapidly classify documents by issue, jurisdiction, or privilege status. Natural language processing can extract entity names, key clauses, and other vital metadata to route documents to the right reviewers.
According to Casey Flaherty, principal at Procertas, AI-based tools can reduce the number of documents requiring human review by 60-90%. This significantly cuts costs and speeds up review timelines. For example, Latham & Watkins turned to predictive coding for a major construction dispute. The AI classified over 80% of 6 million documents as non-relevant or privileged. This allowed Latham"s team to focus manual review on the remaining 20% of high-value documents.
Allens is another firm leaning into AI for document review. They use Neota Logic"s virtual assistant Claire to triage documents and draft privilege logs. Claire accelerated document review by 20-40% across several matters. Allens partner Meena Muthuraman says Claire delivers huge time savings while maintaining quality.
The days of lawyers spending hours poring over legal textbooks and case law databases are numbered. Emerging AI technologies are automating legal research to deliver faster insights with less drudgery. Natural language processing and machine learning are enabling legal research platforms to understand complex questions posed in natural language. Rather than relying on Boolean keywords, lawyers can ask questions conversationally such as "What duty of care does an auditor have when preparing financial statements?" AI can analyze the concepts and dependencies in the question to retrieve the most relevant cases and statutes.
For example, Ross Intelligence leverages IBM Watson technology topower a legal research solution driven by natural language queries. ROSS can quickly reference millions of laws, regulations, cases and documents to compile a customized report on any legal topic. LawGeex takes a similar approach, using NLP algorithms to find answers in seconds rather than hours. Its legal research assistant achieved an average of 94% accuracy across hundreds of commercial legal questions.
Other innovators like Casetext CARA and Judicata combine AI algorithms and legal experts to verify the results. This allows the AI to learn from human researchers and continuously improve its analysis.
The benefits include faster research, more thorough analysis and lower costs. Pinsent Masons adopted ROSS for its teams in Singapore and were impressed by the 70% time savings versus traditional research methods. Clyde & Co saw legal research time drop from 5 hours to 30 minutes after bringing ROSS on board. Lawyers can get quick answers anytime without toggling between online sources or consulting colleagues. The AI becomes an always-on research assistant that frees up time for advising clients.
While AI is automating many legal tasks, seasoned lawyers still reign supreme when it comes to exercising professional judgment. Their years of experience, legal acumen and emotional intelligence give them an edge over machines in high-stakes situations.
Making the right judgment calls requires going beyond rote analysis to weigh nuances and consider potential ramifications. Lawyers must balance legal principles, public policy, ethics, client needs and real-world pragmatism when advising on complex issues. This involves intuition honed from past cases and understanding how certain arguments or strategies may land with particular parties or judges.
Legal judgment also demands empathy, compassion and cultural awareness. A robot may identify the optimal legal argument in the abstract, but human attorneys better appreciate how to tailor their approach in a way that resonates with diverse clients and avoids inflicting unnecessary harm. As Latham & Watkins partner Scott Tseng notes, exercising judgment requires "emotional intelligence " to fully appreciate and advocate for what a client needs and what feels just given the specific facts."
Consider sensitive topics like settling harassment lawsuits. The numbers may indicate an inexpensive settlement makes fiscal sense. However, the trauma inflicted demands deeper consideration of what feels morally just. AI lacks the emotional wisdom to navigate these gray areas. Or take high-profile cases steeped in social issues, like college admission scandals or police brutality cases. The PR optics and potential backlash necessitate human judgment.
Due diligence is a vital but tedious process in mergers, acquisitions, and other major transactions. Lawyers must meticulously review documents, contracts, financial records, IP portfolios, regulatory filings, and other materials to uncover any red flags or deal breakers. This traditionally manual slog bogs down lawyers and racks up huge fees. However, AI assistants are swooping in to transform due diligence from drudgery into efficiency.
"AI can cut due diligence review time by up to 90% compared to human reviewers," explains Casey Flaherty, principal at Procertas. Machine learning algorithms can rapidly analyze thousands of contracts to surface key clauses, extract critical details, and detect anomalies. This allows lawyers to zero in on documents requiring closer scrutiny rather than reading everything themselves. Contract review startup LawGeex touts its AI platform"s ability to "pore through endless piles of contracts" and deliver actionable insights in hours rather than weeks.
Allen & Overy (A&O) recently tested AI tools from Luminance and Kira Systems on a pilot project to analyze acquisition contracts. They found the AI could "quickly read and interpret large volumes of documents" and identify key information faster than lawyers could manually review a small subset. Encouraged by the gains, A&O now uses AI to support due diligence on most major transactions.
Norton Rose Fulbright has also embraced AI assistants for due diligence. They use Luminance to review loan documents across multiple jurisdictions and languages. The AI reviews up to hundreds of documents per hour and flags potential compliance issues. This enables Norton Rose"s lawyers to concentrate on negotiating deal terms and providing strategic advice rather than getting mired in grunt work.
Law firms sit on mountains of data locked away in documents, transcripts, and case files. This data holds immense value for improving legal workflows and unearthing hidden insights. However, manually analyzing massive volumes of unstructured data is impossible. This is where machine learning shines. AI models can rapidly process terabytes of text to detect subtle patterns and derive meaning from the noise.
Several pioneering firms are using natural language processing and predictive modeling to gain strategic advantages. For example, O'Melveny & Myers built a custom tool called Litigation Outcome Prediction (LOP) that analyzes the text of briefs and motions to predict case outcomes. LOP has an accuracy rate of over 80% for binary win/loss predictions. While not a crystal ball, these probabilistic forecasts help O"Melveny assess risks and advise clients on settlement strategy.
Baker McKenzie developed its own AI system called BCourt to forecast enforcement patterns of the European Commission. By analyzing past enforcement actions, BCourt can advise clients on the risks of certain anti-competitive behaviors. Skadden Arps turned to AI to help assess the odds of government agencies allowing mergers. Machine learning models uncover insights from past clearance decisions and microscopic factors that human experts may overlook.
On the document review front, mining millions of pages was once prohibitively expensive. Now predictive coding tools from providers like Logikcull use AI to automatically tag documents by issue, jurisdiction and privilege. Algorithms learn as human reviewers provide feedback, allowing continuous improvement of relevance rankings. This amplifies efficiency and reduces the cost per document reviewed.
Contracts contain a treasure trove of data that holds immense strategic value if it can be efficiently unlocked. However, analyzing contracts at scale has traditionally required armies of lawyers painstakingly reviewing and extracting information line-by-line. Natural language processing finally provides the key to tap into contracts' hidden insights in a scalable way.
NLP algorithms allow machines to "read" unstructured text and convert it into structured data without manual review. This makes lightning-fast analysis of thousands of contracts possible. NLP extracts names, dates, clauses, terms, obligations, limits, and other vital data points from contracts and organizes them into databases. Data analysis and visualization tools can then help lawyers quickly spot trends, anomalies, and relationships to derive strategic insights.
For example, Freshfields Bruckhaus Deringer built an NLP tool called Kira to analyze its transactional contracts. Kira automatically tags key information like governing law, termination rights, liability limits, and change of control triggers. This allowed Freshfields to rapidly benchmark its deal terms against peers to optimize future negotiations. An GC at a professional sports league used Kira to analyze media rights agreements and identify millions in potential new revenue streams within just days.
Allen & Overy (A&O) also embraced NLP to gain a competitive edge in M&A transactions. Its Applied AI team built a custom tool called Luminance to study material contracts. Luminance reads and tags key info like parties, dates, terms, and obligations at superhuman speeds. A&O used Luminance's insights to strengthen negotiating positions and provide better advice on risks and opportunities. One lawyer said Luminance "acted like a junior M&A associate" by unlocking key details from cumbersome contracts.
Clifford Chance created its own AI assistant called MIAC to handle routine analysis of real estate leases. MIAC highlights contract deadlines, termination rights, rent change provisions, and other key terms. This allows lawyers to provide faster service to clients. NLP analysis of lease terms also enabled better portfolio benchmarking and strategic decision making around real estate assets. What once took weeks or months of human effort now happens at machine speed.
Quality training data is the rocket fuel that powers AI"s rise. Machine learning models are only as good as the data they are trained on. Building training datasets requires substantial investments of time, money and human oversight. However, the payoff can be transformative capabilities that evolve through continuous learning.
Leading firms understand the vital importance of fueling AI with rich training data. When Littler Mendelson launched its CaseSmart litigation analytics tool, it hired over 50 lawyers and paralegals to manually review employment lawsuits and label key characteristics. This generated over 500 tags per case including causes of action, outcomes, damages and more. After ingesting this trove of structured data, CaseSmart"s AI could accurately predict the likelihood of success and potential damages for new lawsuits based on the complaint text. The predictive engine keeps learning from new case data, allowing insights to continuously improve.
Dentons also tapped extensive training data to power its NextLaw Labs compliance engine called Compliance.ai. The firm initially hired a dedicated team of lawyers to manually review and tag tens of thousands of SEC comment letters to uncover key patterns and issues. Compliance.ai then used this labeled data to learn how to automatically extract vital details like compliance topics, citations, questions and remedies from raw comment letters at scale. This reduced review costs by up to 90% while still benefiting from lawyers" issue spotting.
When developing BCourt to predict antitrust enforcement, Baker McKenzie likewise relied on practitioners to manually build a database of key details on past European Commission decisions. This provided the necessary training data for BCourt"s algorithms to learn how to predict enforcement outcomes for new cases based on fact patterns. Proper tagging and labeling was essential so the AI understood which details correlated with which enforcement actions.
Never-ending training cycles allow AI tools to keep getting smarter. For example, LawGeex provides lawyers a user-friendly interface to review and correct contract detail extraction by its AI engine. Each correction further improves the NLP algorithms" ability to analyze contracts. This human-in-the-loop approach combines the best of both worlds: machine speed and human judgment.
As AI becomes further entrenched in legal workflows, firms must take steps to ensure these technologies align with professional ethics and client needs. AI comes packed with risks around privacy, security, bias and transparency. However, with thoughtful design and governance, the legal industry can harness AI for good while avoiding potential downsides.
Several organizations have published principles and best practices to guide the ethical deployment of legal AI. For example, the International Association of Young Lawyers issued 10 Ethical Principles for Legal Technology that stress values like competence, confidentiality and transparency. They advise rigorously testing AI systems and providing adequate training so users understand capabilities and limitations. Ongoing human oversight is key to correcting mistakes and preventing harm. The guidelines also emphasize giving clients full visibility into when and how AI is used on their matters.
The UK Centre for Data Ethics and Innovation developed a roadmap for AI auditing and assurance in law. They recommend firms perform pre-deployment risk assessments of AI tools and subject them to continuous monitoring after launch. Documenting data provenance, evaluating bias risks, and monitoring error rates helps to instill public and client trust. Independent audits add further validation, as RightIndem recently conducted for its AI claims assessment tool. The IEEE also released recommendations calling for tech impact statements and controls like ethics boards, diversity mandates and whistleblower policies around legal AI.