Looking Beyond Generative AI – Agentic AI’s Potential in Legal Services
by: Jeff Johnson, Chief Innovation Officer
Generative AI is quickly transforming many industries—including legal services. Many of us are already using (or at least experimenting with) Generative AI, with impressive results. Allow me to introduce the next leg in our journey, Agentic AI. If Generative AI makes our work easier, Agentic AI has the potential to take that up a notch … or ten. There aren’t many Agentic AI tools available yet. It is likely there soon will be. This post is an introductory comparison of how the two differ, relate, and may be used together to elevate our workflows.
What Is Agentic AI?
Think of Agentic AI as a highly capable, autonomous assistant. This category of AI Agents goes beyond simple automation of repetitive tasks (tracking deadlines, monitoring court filings, sending reminders, reviewing a document, etc.). To a degree, Agentic AI can proactively adapt to achieve its defined objectives, without (or with minimal) human involvement. For example, it could automatically notify you if a new court ruling affects your case, provide actionable insights, suggest possible adjustments to legal strategy, or draft an initial response.
Agentic AI is:
- Autonomous: It doesn’t need constant direction.
- Goal-Oriented: It focuses on completing specific tasks or achieving certain objectives.
- Adaptive: It adjusts to changes, like new filings or case developments.
In legal work, this could mean further automating activities in case management, research, drafting, and potentially eDiscovery, leaving attorneys free to focus on higher-level strategic activities.
What Is Generative AI?
Now, let’s talk about Generative AI. This AI type is all about creating content. When you give it a prompt—input context and a question or request—it can generate drafts, summaries, and other written content. Imagine you need to draft a lease agreement—Generative AI can produce a detailed first version based on your input, saving you hours of original drafting time.
Generative AI shines in:
- Drafting initial versions of legal documents.
- Summarizing long documents into quick, digestible insights.
- Tailoring content for different audiences, like clients, courts, or colleagues.
- Generating review decision suggestions (responsiveness, privilege, etc.), with rationale
Generative AI is a creative powerhouse that helps you draft and refine content, documents, etc. Combined with the goal-oriented autonomy of Agentic AI, there is great potential to cover a wide array of activities in both procedural and creative areas of legal work.
What Are the Potential Best Uses of Agentic AI in Legal Services?
Agentic AI, combined with Generative AI capabilities, has the potential to autonomously handle task management, complex content creation, and even decision making.
Commonly cited potential capabilities for Agentic AI include drafting legal documents, summarizing case files, and generating tailored client communications.
The theory is that by integrating Generative and Agentic AI, the system reduces the need for “human in the loop” activities traditionally required in many aspects of legal services, freeing up human resources for higher-level strategic activities.
What are Agentic AI’s Limitations?
Agentic AI does have limitations. It is commonly confused with “AGI” or Artificial General Intelligence. The two differ significantly in scope, capability, and limitations. AGI does not exist today; it is one theoretical future of AI’s evolution in which the AI can perform any intellectual task a human can, learning and adapting across domains, and operating with a level of independence far beyond current Agentic AI capabilities.
Agentic AI automates specific goal-driven tasks, within a relatively narrow focus. It is limited by its lack of general adaptability (beyond that focus), and dependence on human-defined parameters.
What are Agentic AI’s Risks?
As the power and use of Agentic AI evolves, so will the need to address and mitigate the associated risks. Agentic AI solutions, while highly autonomous and efficient, carry significant risks associated with compounding errors and reduced human oversight. Without adequate monitoring, initial errors in decision-making or data processing can propagate through the system, leading to increasingly inaccurate or unintended outcomes.
High-stakes environments, like legal services, amplify these risks. Such unchecked errors could lead to financial loss, failure to meet ethical obligations, regulatory compliance violations, and other significant issues.
For a more detailed discussion of the challenges and risks associated with Agentic AI, I wholeheartedly recommend the following article from the Sedona Conference® written by Tara S. Emory and Maura R. Grossman, J.D., Ph.D., published In The National Law Review (The Next Generation of AI: Here Come the Agents!), available here.
How Might Agentic AI Revolutionize eDiscovery?
Generative AI has truly altered the way we think about eDiscovery document review. We have robust software solutions that harness Generative AI’s large language models and can replicate the productivity of entire attorney review teams, with arguably better consistency, improved accuracy, and decreased cost.
Generative AI enabled eDiscovery review still involves a considerable amount of human involvement in preparing for and validating a review, as well as separate processes for responsiveness review, privilege review, PII and PHI detection/redaction, chronology creation, strategic summaries, etc.
Imagine a world where an Agentic AI solution performs all of the following with minimal human collaboration:
- Utilizes the appropriate combination of search terms, date criteria, email threading, document de-duplication, and an understanding of technical limitations to identify a review population.
- Converts a production request (or similar documentation) into a review protocol optimally drafted for Generative AI enabled responsiveness review.
- Utilizes an evergreen knowledge base of law firms, attorneys, litigation activity, etc. to create an effective protocol for privilege review.
- Utilizes other analytics tools like clustering, entity extraction, and near duplicate identification, to identify population subsets that would be useful in representative pre-review prompt testing.
- Performs, in one pass, the necessary responsiveness review, privilege review, and sensitive data identification.
- Creates a review summary documentation that includes privilege log, chronology of key events, trends, insights, legal strategy recommendations, recommended deposition questions for key custodians, and citations to key documents and/or relevant case law.
Obviously, a system like that represents more of a “wish list” exercise than today’s reality. Even so, I’m confident legal tech product design discussions like that (and beyond) have been happening for years. Agentic AI solutions capable of some portions of that (and similarly complex legal services workflows) will almost certainly exist in the not-too-distant future.
Conclusion
Agentic AI represents another significant evolution in legal technology, complementing and enhancing the capabilities of Generative AI. While Generative AI excels at creating content—such as drafting legal documents, summarizing lengthy texts, and tailoring outputs for specific audiences, Agentic AI can manage task-oriented workflows like tracking deadlines, monitoring filings, and even adapting to new case developments without human involvement. Together, these technologies have the potential to revolutionize legal services, automating complex workflows, reducing the need for “human in the loop” activities, and enabling lawyers to focus on higher-level strategic work. Despite its promise, Agentic AI is not without risks. Compounding errors and reduced oversight could have very costly consequences, highlighting the importance of careful governance. As the legal industry continues to embrace these advancements, the future will likely bring Agentic AI tools capable of handling increasingly complex tasks. Will you be ready?