Mastering the power and peril of AI in legal tech to reach its full potential

AI and legal tech have come a long way, but let’s keep it real. AI is a fascinating field full of endless possibilities, but it’s not all unicorns and rainbows. As we are tracking a technology readiness curve, you can think of AI as a tech journey with its fair share of peak-of-expectation highs and valley-of-disillusionment lows. While some AI technologies have reached a plateau of commonplace, such as OCR for document scanning, others like text generation or even legal text translation still need a human touch to crack their domain-specific complexities.

Meet Maarten, passionate about coding and delivering beautiful, simple solutions to customers. As the Head of Innovation & AI at Henchman, Maarten tries to enrich the lives of lawyers with AI and pushes its boundaries even further. With a useful background as a CTO-turned-Product-&-Tech Advisor for startups, scale-ups (CTO at Showpad, In the Pocket, ..), and investment funds (Miles Ahead, Green Park, ..), Maarten tries to marry product needs with the possibilities offered by technology. Often, this marriage takes a bit of honeymooning to get to the right plateau of productivity.

_{Watch Maarten’s full 20-minute video on AI in legal}

As an illustration of the technology path of AI, Maarten draws inspiration from an MIT report on autonomous driving and its application to legal tech automation. He proposes using a structured framework of levels to navigate the diverse landscape of AI in the legal field. Starting from the fundamentals of document management and search capabilities, progressing through semantic searches, semi-autonomous assistance, domain-specific automation, and ambitiously striving for fully autonomous workflows, each level represents progress in leveraging technology to enhance legal processes.

• Level 0: No Automation
  Picture this: legal professionals huddled around dimly lit offices, drowning in mountains of paper, desperately clinging to their typewriters. Luckily, most – if not all – lawyers have progressed beyond this stage, with digital tools and document management systems now as commonplace as the office coffee machine.
• Level 1: Simple Assistance
  Get ready for some AI magic! At this level, AI technologies assist with tasks like searching contracts and locating specific clauses within a document. This is needle-in-a-haystack at its most fundamental: getting out the data that answers a request, and is still an active problem. How often have you tried looking for a contract that was written before, based on a vague set of details?
• Level 2: Basic Natural Language Processing (NLP)
  At this stage, AI technologies employ semantic searches and analyze textual content to derive classifications and meanings, enabling more advanced searching capabilities. It’s like having an AI Sherlock Holmes equipped with a magnifying glass for your searching needs. This is a place where generative AI already fills in a new set of capabilities, which fundamentally generates answers based on questions instead of simply retrieving the right data.
• Level 3: Semi-Autonomous Assistance
  This level focuses on creating a knowledge base that exceeds simple search and retrieval. One key example is playbook automation to master contract negotiation and drafting based on your historical scenarios – and for instance the client’s redlines. Another example is the AI suggesting alternative or added phrases when you’re writing a clause, without actively inserting itself in your workflow.
• Level 4: Domain-Specific Full Automation
  Consider a highway that has a lane reserved for self-driving cars. In a similar way, only a part of your legal domain could be covered by a more autonomous assistant. By focusing on specific domains, like M&A transactions, AI can offer fully autonomous solutions tailored to the unique challenges of each domain. For instance, when it comes to risk management in a purchase agreement, an AI system that is more familiar with the typical standards and sensibilities involved can autonomously create the right ingredients of legal clauses based on the case. There is a wide gap between Level 3 and Level 4 where a lot of this assistance can still be offered semi-autonomously.
• Level 5: Moonshot Vision
  At this level, we’re dreaming big and aiming for the stars. It’s about the allure of achieving fully autonomous workflows in legal tech. This allows lawyers to focus on the services provided more fully and automates away the actual administrative efforts. However, it is good to remember that a Level 5 automation is better served as a horizon than the actual goal that needs to be attained. A fully autonomous solution consists of various parts that each have their own technology readiness curve, so it’s not always clear when you’ve ​“got there” and often there is much more value in the path than in the destination.
  While generative models like language models can sprinkle their magic on text generation and concept exploration, they should be combined with domain-specific knowledge and databases to maximize their usefulness in the legal domain. Think of them as trusty sidekicks to lawyers, not replacements. These tools can be valuable aids in drafting, exploration, and typing assistance.

Taking it one step at a time is crucial, while recognizing the value provided by intermediate solutions. While generative models like GPT and its chat-based or other API-based interfaces hold great potential, they can also be prone to generating inaccurate or hallucinatory information. Careful consideration is necessary to ensure that legal concepts, integrity, and accuracy are maintained when leveraging these tools.

Those with proficient prompting skills may gain an advantage over others

As we enter this exciting era of AI and legal technology, those with proficient prompting skills may gain an advantage in accessing relevant information efficiently. However, it is essential to acknowledge that relying solely on these tools may still be limited to certain cases. A cautious and informed approach that combines human expertise with AI capabilities will pave the way for a more peaceful and productive integration of technology in the legal field.

Remember, the ultimate goal is not just the moonshot of fully autonomous systems, but also the tangible value created through each step of the automation journey. By leveraging AI in a domain-specific manner and continuously evolving our understanding of its capabilities, we can navigate the complexities of AI in legal tech and harness its true potential to empower and enhance the work of lawyers.