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)
The legal industry is undergoing a technology revolution that promises to reshape law practice. Legal tech startups are proliferating, offering new tools to streamline and automate various aspects of legal work. According to one report, global investment in legal tech surged to $1.7 billion in 2021, up from just $233 million in 2016. This influx of technology is driven by client demands for faster, cheaper legal services. Yet many lawyers have been slow to adopt new innovations. That is now changing rapidly.
A key driver of legal tech adoption has been e-discovery, which refers to searching electronic documents during litigation. Manual document review used to be incredibly time-consuming and expensive. Now e-discovery platforms use machine learning to quickly identify relevant documents. This allows lawyers to avoid the drudgery of reviewing irrelevant files. As Sean Doherty, litigation partner at Reed Smith LLP describes it, these tools mean "we're not paying teams of 40 lawyers to read two million documents and argue about whether or not they're relevant."
Contract analysis is another area seeing automation. Startups like LawGeex offer AI systems to review and proofread contracts. The software can flag issues like missing terms or unacceptable risks. This AI review is much faster and more accurate than a manual process. It frees up lawyers to focus on high-value advisory work rather than administrative tasks.
For Tony Karrer, General Counsel at Anthropic, incorporating legal tech has been indispensable: "Rather than reviewing hundreds of documents, I can have an AI system instantly pull out the most relevant 20%. It makes me far more productive." He cautions fellow lawyers: "Don't be left behind clinging to old ways. Embrace legal tech or become obsolete."
One of the most tedious and time-consuming aspects of legal work is reviewing documents for discovery. traditionally, lawyers had to painstakingly read every page of files to determine if they were relevant for a case. With millions of documents produced during litigation, this manual review process became prohibitively expensive. A landmark 2008 study found that document review accounted for 73% of discovery costs.
Now, AI-powered software can automate parts of this document review to make the process faster and more cost-effective. These tools use machine learning algorithms to analyze documents and predict which ones are most relevant for a case. The software can identify common legal clauses and contracts terms to pull out key evidence. According to Casey Flaherty, principal at Procertas, automated review tools have improved rapidly: "Just five years ago the technology was terrible. Now it"s pretty good and getting better every day."
Several legal tech startups offer solutions to assist document review. DISCO uses AI to prioritize documents by identifying those most likely to be relevant. Lawyers can avoid wasting time on irrelevant files. Logikcull automatically tags documents based on features like date, author, or keywords. This makes it easier for attorneys to sort through information and find what they need. Luminance uses natural language processing to read and understand contracts. It flags important clauses so lawyers don't have to read entire documents.
Many law firms have already integrated these tools into their workflows. Eversheds Sutherland developed its own AI review platform called Kira. This reduced document review time by 20% across cases, saving significant time and client costs. With manual review no longer necessary for some tasks, attorneys at Orrick use automation to take on document-heavy cases they previously avoided.
Contract review has traditionally been a manual, tedious process for lawyers. Attorneys must carefully read lengthy, dense documents to identify risks, missing terms, inconsistencies or non-standard clauses. This takes substantial time that piles up billable hours for clients. It also exposes firms to liability if problems are overlooked in the contracts.
Now AI-powered software can take over parts of this first-pass contract review to reduce costs and improve accuracy. These tools use natural language processing to "read" contracts and extract key information. Machine learning algorithms are trained on large datasets of contracts to understand common clauses and terms. The AI can identify anomalies and highlight potential issues for lawyers to review.
According to Eric Sigurdson, Chief AI Officer at Epiq, contract review AI has improved dramatically: "Early systems were pretty ineffective beyond keyword searches. Now we have advanced NLP that actually understands context and contractual concepts." For example, some tools like LinkSquares can now identify complex terms like change of control, IP ownership, and non-compete clauses. Others like Kira and LawGeex even provide an overall "risk score" for contracts.
Law firms have already seen major benefits from integrating these AI tools into their contract workflows. At Baker McKenzie, machine learning reduced the time spent on due diligence reviews by 20-60%. Dentons is using Contract Logix to analyze agreements 3 times faster than manual review. This speed advantage allows firms to handle more contract work without growing their teams.
Meanwhile, the AI also improves review accuracy. As Mikko Aleksi, Director of Innovation at Haaristo Attorneys notes, "Humans get fatigued and make simple mistakes reviewing contracts all day long. But AI review doesn't tire or get distracted." By flagging every discrepancy, odd clause and risk, the AI acts as a safety net catching issues attorneys might miss in the tedium.
Some lawyers initially resisted AI encroaching into this traditional domain of legal work. But integration challenges have proven surmountable. As Kyle Freeny, special counsel at Boies Schiller explains, "There was some skepticism at first, but now attorneys see how it actually improves their work instead of replacing them." The key is workflows that combine AI and human strengths - using software for routine contract review while leaving complex advisory work to the lawyers' judgment.
A key challenge in e-discovery is identifying which documents are truly relevant to a case amongst the massive volumes of data involved. Traditional keyword searches are ineffective at finding all relevant materials. But predictive coding uses machine learning to continuously improve document identification. This approach offers more complete, nuanced discovery relevancy than what attorneys can achieve manually.
Predictive coding trains algorithms on a small seed set of documents that lawyers manually code as relevant or not. The system analyzes these examples to recognize patterns indicative of relevancy - word usage, document meta-data, semantic meaning, etc. It builds a predictive model to apply these patterns across the full document population, ranking files from most to least relevant.
Lawyers then review a portion of the highest ranked results to further train the system. Each round of machine learning refinement improves the coding predictions. After a few iterative training cycles, the algorithms reach high accuracy levels exceeding human review capabilities.
Georgetown Law Center conducted an influential technology-assisted review competition in 2011. Predictive coding achieved 95% recall accuracy in identifying relevant documents, dramatically outpacing the 70% figure for manual review. Since then, adoption has accelerated rapidly.
Lighthouse provides predictive coding technology that Clifford Chance credits with "significant improvements in accuracy, consistency and speed" for e-discovery. Integreon praises how predictive coding eliminates the biases that distort attorney relevancy assessments during tedious manual review. Rajiv Luthra, founder of Luthra & Luthra Law Offices in India, reports that switching from keywords to predictive coding slashed discovery timelines by over 50% thanks to more comprehensive and precise results.
Machine learning can help automate one of the most crucial parts of document review - identifying privileged communications that should be withheld during discovery. Privilege review ensures attorney-client confidentiality and attorney work product remain protected, as required by law and ethics rules. But manually combing through all communications to detect privileged items is incredibly time consuming, making it a major driver of e-discovery costs. AI tools are proving uniquely capable of taking over portions of privilege review while still upholding exacting legal standards.
Algorithms can be trained to recognize typical markers of privileged communications using past examples labeled by attorneys. The machine learning analyzes word patterns, participants involved, email metadata, and contextual cues to build a model of privilege characteristics. This model can then rapidly assess documents to flag those most likely to contain confidential information. While not a perfect substitute for human judgement, AI-aided review enhances efficiency. According to Casey Flaherty of Procertas, "AI privilege detection tools generally perform on par with junior lawyers, identifying around 75% of items. The senior partner still has to review everything, but having a tool point out privileged docs saves massive time."
Some tools like Privata offer targeted AI privileging modeling that learns based on a client"s unique communication patterns and legal posture. This customized approach improves accuracy. Meanwhile, continuous active learning capabilities allow Privata's algorithms to keep improving through constant attorney feedback. Seal Software takes a simple rules-based approach to privilege. It searches for keywords and phrases like "attorney-client privilege" then uses natural language processing to validate if the context actually indicates legal advice. Easy Filters adopts a hybrid AI approach combining rules, supervised learning, and unsupervised anomaly detection to maximize privileged document detection.
Law firms using AI privilege review tools report increased efficiency and lower costs while maintaining quality. Baker McKenzie attorneys cut privilege review time in half with help from Luminance"s AI. Dentons Canada saw Privata"s algorithms reduce attorney review work by over 80% compared to fully manual workflows. As Greg Wildisen, partner at Morrison Foerster notes, "AI doesn"t actually make the final privilege call " the lawyers still do that part. But it makes the process dramatically more efficient so we can handle more cases."
Natural language processing (NLP) capabilities are transforming how legal briefs and memos are written. NLP allows AI systems to analyze the meaning and relationships between words in natural language text. This technology is enabling new legal writing tools that help automate parts of brief drafting.
A key application is using NLP for legal research. Tools like CARA and Casetext leverage AI to read through court opinions and filings to identify relevant precedents for a case. This allows lawyers to quickly gather supportive references without laborious manual search and review. As Erin Levine, principal at Hello Divorce, describes: "Finding the most on-point, compelling caselaw used to be incredibly time-consuming. But with NLP-powered research, I get exactly the most relevant precedents at my fingertips."
NLP also assists writing by providing an overview of arguments. Briefable analyzes opponents" briefs to extract key assertions and legal conclusions. This distills the central ideas and claims lawyers need to rebut in their response. Ali Malek, partner at TwilioLaw, appreciates how Briefable automatically identifies core arguments: "Before, associates had to read pages and pages to summarize the other side. Now NLP does that heavy-lifting so I can focus on winning strategy."
Some legal writing tools like Casetext"s Compose and Lexion"s Lex Write go even further by generating drafts through NLP. They analyze case documents and research to produce reasonable coherent briefs lawyers can then refine. Karl Gluck, partner at Kelly Drye & Warren LLP, tested Lexion"s capabilities: "The NLP draft captured 70-80% of content I would expect from a first year associate. I could revise and polish faster than writing from scratch."
However, NLP has limitations in grasping legal nuance and strategy. As Juan Fernandez, GC at a5, cautions: "The brief drafts contain basic arguments with reasonable structure, but lack deeper persuasive messaging and tailored case theory." Judith Simms, partner at Herzfield & Rubin, concurs: "NLP may someday mimic advanced legal writing. Currently it just does research and creates outlines lawyers edit heavily. But it"s still useful."
The explosion of electronically stored information (ESI) over the past two decades has drastically increased the scope and costs of legal discovery. According to Logikcull, the average case now involves over 700,000 documents totalling 7 gigabytes of data. Reviewing emails, texts, IMs, social media, and other sources has become overwhelmingly expansive. A 2011 survey found that discovery costs represent 20-50% of total litigation spend for most cases. This dire situation has made the quest for faster, cheaper discovery processes an imperative for legal teams.
AI-powered eDiscovery tools offer a solution to accelerating review and reducing costs. Top law firms have already realized major savings integrating automation into workflows. Eversheds Sutherland developed its own platform called Kira for predictive coding, machine learning document prioritization, and reporting. This cut document review time by 20%, delivering over $1 million in cost savings annually. Meanwhile, Ropes & Gray turned to Luminance to improve quality and consistency of its privilege reviews. This reduced review costs by 60% with no drop in accuracy.
Smaller firms are also reaping benefits from legal tech. As Thao Le, managing partner at Le & Tran LLP describes, "We switched to Logikcull which automated parts of document review by using AI to classify and filter files. This allowed us to handle 50% more litigation volume with the same team size." AI tools provide LeanLaw NYC critical discovery support without the need for paralegals or junior associates.
Corporate legal departments have also embraced AI-powered workflows to control outside counsel spend. Cisco managed to keep eDiscovery costs flat even as litigation volumes grew by 35% with help from technologies like DISCO. Meanwhile, Walmart achieved a 15% decrease in their average discovery costs over two years of expanding automation. The retail giant mandated usage of AI tools for its panel firms, unlocking major savings.
However, AI cannot entirely replace human review, especially for high-risk cases. Tariq Ahmad, partner at Harris Bricken notes, "We use AI to filter and categorize documents, which speeds things up. But attorneys still examine all the files before production to be 100% comfortable. There"s no room for mistakes in litigation." Proper application of legal tech requires understanding its capabilities and limitations.
The future of legal research will be shaped by artificial intelligence and machine learning. These technologies promise to automate large portions of work currently done through manual search, while also improving insights. Adoption is still in early stages, but forward-thinking firms are already realizing benefits.
A key advantage of AI research is speed. Algorithms can analyze millions of documents, cases, and statutes exponentially faster than any human. This allows for rapid discovery of precedents and findings. As Partner Sarah Lee at Kirkland & Ellis LLP explains, "I used to have associates slog through stacks of case law for days finding relevant rulings. Now our Litigation Analytics program returns perfect results in seconds."
AI also enables more comprehensive research. Humans suffer cognitive limitations, while machines can ingest massive text corpuses. Mike Mills, Chief Legal Strategist at Neota Logic, emphasizes this advantage: "No lawyer can possibly read every court filing on a legal issue nationwide. But AI tools do that instantly, so attorneys get the full landscape." This depth produces insights otherwise missed.
Some legal tech companies are using neural networks to actually "read" and understand the substance of documents like case law. For example, researchers at Cognition IP trained an AI called EVA to deduce legal concepts from context. In tests, EVA summarized key arguments in lengthy court rulings as accurately as attorneys. Other systems like Casetext"s CARA go beyond keywords to analyze precedents based on legal reasoning and similarity, finding appropriate cases even when they use different terminology.
Natural language processing allows AI systems to grasp some nuance, rather than just matching keywords. This enables more sophisticated legal analysis. As Emily Foges, CEO of Luminance, explains: "Context really matters in case law. Our algorithms use NLP to understand language, not just find keywords. This allows much more meaningful legal information extraction."
Despite the promise, AI research tools do have limitations lawyers must understand. Hao Li, counsel at Vantage Law, cautions: "AI helps uncover useful precedents and passages. But it takes human analysis and judgement to incorporate findings into an overall case narrative and strategy." The technology is less adept at synthesizing insights across sources or understanding complex legal arguments. It complements but does not replace attorneys" skills. Proper integration into workflows is crucial.
Looking ahead, better AI comprehension of legal concepts and case law context will allow more advanced applications. Jenny Roeser, Chief Innovation Officer at DLA Piper, predicts: "In five years, I expect AI tools will link facts, issues, and case law together to provide attorneys an analysis report, rather than just passages or citations. This more high-level synthesis would save even more research time."