Legal AI is rapidly becoming a killer use case for LLMs. You could even argue LLMs were born to take over the legal profession. Legal tasks are already being accelerated and automated, and we see a clear opportunity for generational impact on the lives of lawyers. But what would it take to build AI lawyers?

In this piece, we highlight the significant implications for the industry, the current market gaps, and ultimately, what it will take to achieve AI-powered lawyers.

Globally, lawyers review up to a billion unstructured documents daily, requiring a massive amount of understanding to advise their clients effectively on the law. To make this even more challenging, the number of these documents increases and changes regularly as new cases emerge, making the knowledge base dynamic and not static.

Before LLMs, the challenge of synthesising, drafting, and understanding these large volumes of text-based information was immense. Lexicode-based search and NLP-based solutions were only able to tackle certain particular use cases after careful training by data scientists. Generalisable semantic understanding of large contexts was a distant prospect.  A legal associate on the team is often referred to as the “Wikipedia” after days or weeks of deep analysis of thousands of documents. 

Now, the ability for automated results has taken a giant leap for lawyers, and possibly humankind. The volumes of content can finally be understood and leveraged by LLMs as model context windows grow, inference costs shrink, and the ability for semantic search with accuracy arises from a combination of keyword, vector, and graph databases. 

The ability to craft and refine expert-level content is also delivered through careful context engineering around the use of models and fine-tuned models, trained using reinforcement learning with human feedback.

New technology solutions no longer only add value for tactical and repetitive use cases; they are also used for complex and strategic legal work.

LLMs and the tooling stack around them are changing how lawyers work. The core generative AI use cases of research, summarisation, and drafting have historically accounted for two-thirds of lawyers’ working hours (source). Through its understanding of unstructured texts, AI can speed up these labour-intensive tasks. In recent years, legal content platforms and tech incumbents have invested heavily in AI. For example, Thomson Reuters acquired Case Text for $650m (source) almost as soon as ChatGPT launched in a defensive move against LLMs. On the administrative side, Wolters Kluwer recently announced its acquisition of the AI-powered legal billing solution Brightflag for €425m (source). Several AI-native contenders have also emerged, including Harvey AI, which is reportedly raising another $250m (source).

The AI opportunity at stake here is tremendous. The legal market generates approximately $1 trillion in annual revenue. With such a strong tech-market fit, the legal opportunity for efficiency alone is worth up to $300 billion. We know that two-thirds of legal time is spent on AI-relevant use cases, and our conversations with law firms indicate an efficiency gain of up to 50% in the near term. Even if we assume only a 25% efficiency gain, this still represents a $150 billion efficiency opportunity. Beyond this, there may be an even greater revenue opportunity unlocked by generating faster, sharper and more relevant insights. 

However, there is a reason why we estimate the near-term efficiency opportunity for AI to be 50%, not 100%. To further digitise the legal stack, we will need to see deep context and precision, as well as end-to-end workflows for specific use cases that drive ROI. In fact, we expect to see generalist horizontal platforms paired with increasingly specialised products for particular practice areas. Beyond solutions that automate legal tasks end-to-end, digitisation will extend to client administration and project management. And, the legal industry may need to shift how it charges – an unimaginable revolution. 

Once you can do it all with AI, the law firm and the in-house legal department fundamentally change. But this is a hard, high-precision problem. 

To break down the problem, we think it’s helpful to consider how lawyers spend their time today. Time is the unit of account in law, and technology is compressing it. Legal teams spend their time across five tasks, which AI is accelerating:

• Drafting and reviewing: Generating legal documents and ensuring accuracy and precision for their required purpose – the bread and butter of lawyer output. We estimate a 30-40% time-saving opportunity here achieved with a high-quality AI-driven first pass.
• Research: Finding answers to questions on regulations, historical precedents, and expected future changes to law, often core to delivering on a firm’s “expertise” value proposition. Given the significant volume of documents lawyers typically have to sift through, they are reporting 50-60% time savings, with lawyers carrying out the final mile of sanity checks.
• Due diligence: Scanning, summarising, and identifying risks in hundreds of documents for important deals or transactions – often high-volume paralegal work – can be substantially automated with 50-60% efficiency improvements.
• Project management: Navigating emails, document versioning, and stakeholder coordination, which all take far longer than lawyers would like to admit. We estimate AI-supported project management can accelerate management by 30-40%.
• Client admin: Client onboarding and offboarding, including KYB, engagement set-up, and billing – ideally led by support staff, but often taking up lawyer time. We see 60-70% efficiency opportunities for retail law due to the high volume of standardised client onboardings, and 20-30% improvement for commercial law onboarding due to higher complexity and nuance.

Across these, we see legal teams increasingly able to reduce hours on non-billable work, whilst increasing capacity for billable hours and more strategic, high-value work.

We see the market coalescing around several tech solutions:

• Horizontal copilots: Creating the central go-to AI solution for legal teams – these are the juggernauts of the legal AI world, serving everyone in the market, from SMBs up to the largest enterprise firms.
• Practice area suites: Bundling high-quality, end-to-end solutions to drive deep productivity improvements for specialised use cases such as IP, litigation, real estate or certain types of M&A transaction work.
• Legal management software: Software to manage the heart of the law firm, including (1) practice management, with CRM, ERP and document management functionalities to maximise ease of access and value of internal content, and (2) operations software, supporting the bookends of legal operations such as onboarding, billing, risk and more.
• AI-first law firms: Offering fast, high-quality legal services directly to clients by combining the latest technology and models with AI-native lawyers who know how to use it best.

External content provider incumbents will also remain key players in the landscape. Holding rich libraries of case data and regulatory content, they remain critical enablers of high-quality, case-relevant, and precise legal automation and insights.

There are still several friction points for the adoption of legaltech today. Despite strong market pull with law firms and in-house lawyers now testing and selecting AI vendors and in-house solutions, only 30% of lawyers are actually using AI (source). Widespread adoption is hindered by unresolved tasks and use cases, a lack of trust and auditability, and limited tolerance for change in the legal profession.

Unsolved workflows: Legal AI copilots alone are insufficient to drive scalable adoption. To go beyond personal productivity and be used collaboratively, AI assistants must integrate bespoke workflows that support lawyers’ work end-to-end.  

This is driving a convergence of original “workflow builder” players such as Bryter, who have released AI assistants, and the “AI assistant” players such as Harvey and Legora, who are releasing basic workflow builders. There is a powerful opportunity to put AI into a customised workflow to create and automate work at scale.

Specialised use cases: Early movers in the legal AI space, like Harvey and Legora, focused heavily on horizontal solutions. Adopting AI was a legal imperative for law firms, and as such, the result was the widespread adoption of horizontal use; however, many practice-specific use cases remained unsolved. Buyers are increasingly realising that fine-tuned solutions for their specialist practice areas, such as IP or Real Estate, with deep integrations and differentiated content, will create lasting value. 

Precision anxiety: Lawyers need confidence and trust in their tools, especially in a world in which one hallucination can sink a deal or a case. With current tools, lawyers have to always check, review, and complete the last mile. Trust takes time, and lawyers trust precedent and practice, not flashy software releases. They need to see high-quality, firm-grade performance, with auditable and deterministic outcomes. In other words, AI must consistently arrive at the correct answer and demonstrate its reasoning process.

Deeply ingrained habits: In an artisanal and academic industry, things have been done in certain ways for a long time. Habits are notoriously difficult to change, even in the best of circumstances, and effective change management is necessary. Integrations with core tooling, such as iManage, NetDocs, and Teams, are now table stakes, and the UI must feel familiar.

The billable-hour paradox: For many firms, even if they are confident in the value of the AI, they are trapped by their pricing model. More than 60% of firms still charge by the hour (source), rather than by legal value produced, which does not incentivise efficiency. We can expect to see increasing experimentation with alternative fee arrangements, and AI-first law firms, such as LawHive and Avantia Law are also well-positioned to capitalise on this revolutionary shift.

Picture a world where a client uploads a request, including key documents such as pre-existing agreements and dispute information. Ten minutes later, they receive a risk-scored action list to close their transaction, which is partner-verified, regulator-compliant, and pre-priced.

In that world, law firms are product companies, and the billable hour is a legacy metric. Accurate legal work is table stakes, and they compete on client experience: speed, ease of use, and quality of strategic insight.

There’s yet a long road to replacing lawyers in the law firms and in-house counsels of the future – but we believe the big winners in legal technology will be those who can solve for high precision outputs, direct ROI impact, end-to-end legal operations, or drive a shift in business model.

First and foremost, legal AI applications need to demonstrate high precision and relevance by integrating specific content. A key battleground here is integrating firm-relevant, case-relevant, and industry-relevant data from content providers. Both generalist and specialist practice solutions will enable seamless connectivity with proprietary datasets within the firm, as well as built-in access to external, industry- and country-specific content. 

This is already happening, as underscored by two recent moves. First, Harvey’s recent move to partner with LexisNexis (source), a leading global source of case law and regulatory data (and also an investor in Harvey). This will be a game changer for Harvey’s content quality and relevance. Similarly, Clio’s acquisition of vLex for $1b (source) has expanded their AI capabilities with a deep well of data on statutes and case law.

Secondly, applications need to go beyond “chat GPT-like” features into workflows to drive firm-wide ROI. This trend will be enabled by both horizontal legal platforms and practice-specific solutions, much like in other sectors. For example, banks purchase both generalist AI platforms (e.g., Dataiku) that drive the adoption of AI, ML, and data science across the enterprise, as well as verticalised AI vendors (e.g., Quantexa) that support decision intelligence with unique capabilities for financial services in areas such as KYC, AML, risk and fincrime. 

In legal, Bryter is an example of a generalist horizontal solution with deep workflow-building capabilities and integrations to allow AI to become part of any customised legal workflow. We believe generalist horizontal platforms like Byrter, Harvey, and Legora will continue to build workflows to support transactional work for any lawyers. At the same time, some will focus specifically on in-house lawyers (e.g., Wordsmith, Legalfly, Flank AI) or “door law” for SMBs and retail clients (e.g LexRoom AI, Jupus).

On the other hand, we believe there is also an increasing opportunity for new, focused, practice-specific legal suites.

We see significant opportunities across various areas, including IP (e.g., Lightbringer, Ankar), Litigation (e.g., Wexler), Real Estate (e.g., Orbital, Conveyd), and other domains.

Thirdly, to achieve the dream of ‘AI lawyers’, solutions will be needed to solve for end-to-end legal operations. AI drafting is cool, but it’s not all. There are nitty gritty pain points across the journey, including client onboarding, KYB, project management, billing and settlement, document handling and sorting, larger scale document creation or re-papering, filings, regulatory assessments, approvals, and more. The full stack must integrate with existing tooling and, have the deepest security and governance to handle the most sensitive cases. These operations may be served by broader horizontal solutions or by legal-specific products that stitch together the needs for a specific legal use cases end-to-end.

Finally, the winning legal services won’t always be the incumbents. The winners will be those who can shift the legal business model to capture the value from increased efficiency. AI-first law firms have an advantage in this race, especially when they can offer a comprehensive client solution priced based on outputs rather than billable hours. We are already seeing this delivered in different approaches, from marketplaces (such as Lawhive) to those managed with account-based relationships (such as Avantia). Data-driven workflow optimisation, and strategic applications of AI to the highest-leverage use cases, will drive faster, higher-quality, and more scalable solutions.

In the end, it is a hard problem to replace a lawyer – although client-facing AI lawyers are becoming possible, nobody wants an AI lawyer that makes mistakes. The industry needs to be risk-averse. We expect human lawyers to remain relevant. Indeed, they are even becoming more high-value and strategic for their clients when supported by AI, as the operational work is automated away.

Nevertheless, as we increasingly see glimpses of a possible future run by AI lawyers, it raises philosophical questions. Would our legal systems run better on AI? Do we want AI litigators and negotiators? AI judges? Or even, AI jurors? For the first time, these questions are within our reach.

If you’re breaking barriers in legal tech, or other AI-enabled services, please reach out. We’re excited about the AI-enabled services opportunity in legal, and across accounting, fund admin, tax advisory, consulting, compliance and more.

You can find us at shamillah@dawncapital.com, andrea@dawncapital.com, zoe@dawncapital.com and evgenia@dawncapital.com.