Artificial intelligence (AI) has developed significantly from an innovative tool into a transformative entity in numerous industries, including the legal profession. For many lawyers, AI may still be associated with predictive analytics or document review tools. However, AI development has taken major leaps. Agentic AI has moved from concept to horizon.
Unlike generative AI (GenAI), which produces content based on prompts, agentic AI takes initiative. It not only interprets instructions but can also execute them across various tools and platforms autonomously. This advancement represents a notable change in how legal professionals may engage with technology. Understanding agentic AI is now essential for firms, legal departments and legal aid organizations.
Agentic AI is in its infancy. Agentic AI differs from GenAI. GenAI uses advanced algorithms to create new content, including text, images, videos and code. Agentic AI features autonomous AI agents that learn and adapt to do specific tasks. The purpose of Agentic AI is to perform tasks that humans would typically do, including following a set of instructions and executing actions. The technology differs from automation tools like Zapier because it can make decisions and adapt to changing conditions.
Given access to a set of tools, AI agents can make decisions and accomplish a specific goal with limited supervision. For instance, Jennifer Case notes in her LawSites article “The Battle for Small Law Dominance in the AI Agent Era: Microsoft vs. Google,” an AI agent could monitor court dockets and calculate deadlines, alerting you when opposing counsel files a response or help with discovery review or draft LinkedIn posts based on recent news.
Agentic AI promises to reduce the need for people to do mundane, repetitive or low-effort tasks. That said, agents will still need to be managed, and their work will need to be checked for accuracy and completion. Managers will need to shift trust from humans to algorithms.
Joe Patrice at Above the Law asks the question that no doubt occurs to many lawyers: do lawyers actually want an agent? In reviewing Thomson Reuters’ agentic evolution of CoCounsel, he opines that in an industry that routinely marks down work conducted by elite law school graduates, it is not clear that lawyers would want an algorithm to do the work.
The promise of sgentic AI has not yet been fully realized, and the technology is far from fulfilling the potential of being truly autonomous. It will be increasingly important to ensure that the “human in the loop” oversight is emphasized when deploying agentic AI.
If you read about what agentic AI can and will do and see opportunities for security, ethical and privacy concerns, you are not alone. “Signal President Meredith Whittaker calls out agentic AI as having 'profound' security and privacy issues,” noting that the access these bots would need to have in order to fulfill their tasks would break the “blood-brain” barrier between the application layer and operating system layer, creating impossibilities in existing protections to keep information protected. For agentic AI to work you would need to give it access to multiple tools to get the job done. Examples like letting an agent book a flight for you, plan a trip or buy concert tickets would require you to give access to multiple tools that house your private information and use your credit cards on your behalf. That is a lot of trust.
Agentic AI is a scammers' paradise. They can deploy agents at scale to make mischief. With a combination of agentic and GenAI, social engineering–based scams have substantially accelerated. With access to interests, hobbies and even your likeness easily accessible, an agentic bot with deep-fake capabilities can target individuals and hold conversations that mimic human interactions. Security experts are suggesting that people have a safe phrase shared with their family and company to confirm they are talking to them. Clients should be on that list, too.
In a legal context, attorneys at Proskauer Rose investigate the hazards of agentic AI in contract law, asking “who is really clicking ‘accept’?” For instance, is a transaction initiated and executed by an AI tool on behalf of a user enforceable? If a dispute arises over an e-commerce transaction are current laws enough to protect the parties, or could they exacerbate difficulties in resolving disputes when an agentic AI tool made the decision on behalf of the user? Lawyers and their clients will need to understand the implications of AI developer terms of service. The authors summarize that decades of established law may now be challenged by agentic AI.
Another aspect to consider with agentic AI is ethics. If an agent is given autonomy on a law firm’s website with no oversight to answer questions posed by potential clients, who is responsible for inadequate or incorrect information? In consumer protection, will agents making decisions for companies and products introduce bias that harms or discriminates against people? MIT’s Moral Machine has long been a source of studying moral decisions made by machine intelligence, evidencing how human bias can influence autonomous technology like self-driving cars.
How will lawyers use agentic AI in law practice? Here are some examples:
With companies and large law firms deploying agentic AI–based services, what tools can law firms, in-house legal departments, legal aid organizations and other legal services providers begin to investigate for internal use and externally facing legal help? Here are some of the legal and business-oriented agentic AI products rolling out to the marketplace. These tools should be considered in the “emergent” phase, with their full potential more conceptual than reality.
LexisNexis has launched Protégé AI Assistant. This product will replace the AI Assistant on Lexis+ AI, and Protégé will be ubiquitous across all LexisNexis products. Protégé’s agentic AI will autonomously complete tasks such as suggesting workflow actions based on the type of document uploaded, drafting deposition questions and discovery documents, generating timelines, analyzing transactional documents and more.
Thomson Reuters is rolling out agentic AI across CoCounsel products, leveraging the acquisition of Materia, an agentic AI for the tax, audit and accounting professions. Agentic capabilities for CoCounsel for legal will roll out in the summer. The upcoming release is expected to include workflows for drafting, employment policy generation, deposition analysis and compliance risk assessments.
Definely, a contract drafting and review company, is offering Enhance, an agentic AI legal assistant integrated into Microsoft Word. The agent will draw information from the suite of Definely tools, including document intelligence, precedent clause bank and an automated proofreading tool.
Harvey AI is rolling out AI Agents to help build workflows. Harvey agents will guide users through steps in a task and reduce the need to craft detailed prompt queries. Custom evaluations will check the agent’s work against human quality work on common tasks.
Microsoft 365 Copilot Wave 2 was announced at Microsoft Build. These tools will build out the existing Copilot Agents including Researcher and Analyst. Researcher is a “supercharged” version of Deep Research. Analyst helps format your data to help it make sense. Other tools that Microsoft is rolling out include Copilot Notebooks that turn Copilot output into an interactive document. Eventually, the plan for Copilot will include the ability to control your PC with app actions. Current subscribers to Microsoft 365 Copilot for Business can check out Researcher, Analyst, Visual Creator, Writing Coach or create an agent in the Copilot interface.
Jotform, the survey tool, has rolled out AI Agents for customer service. AI Agents in Jotform can collect data, answer user questions and provide personalized guidance. It can be used as frontline customer service, with phone agents, live chat and more. There are plenty of templates ready to customize and deploy including some that law firms might be able to use, like appointment and booking AI agents, human resource AI agents, customer service AI agents, satisfaction survey AI agents and intake AI agents. There are already legal-specific AI agent templates including intake AI agents, legal consent AI agents, Legal Aid Application AI agents and many practice area specific AI agents.
Zapier Agents will give users the ability to organize, monitor and launch fully autonomous AI workflows. Zapier is already a powerhouse tool for connecting applications to create workflows. Zapier agents will tap directly into nearly 8,000 apps to act on your behalf across your entire tech stack with no coding. You can organize Zapier Agents into “pods” to organize related agents into logical groups like “content marketing” or “client intake.” There is an activity dashboard so you can check on your agents to see what they are doing and if any need attention. Of course, there are ready-made templates, like the Google Business Review Responder.
Google announced further AI development at the recent Google I/O 2025 conference, including a low-code agent builder for creating custom AI assistants and enhanced Workspace “gems” that work with files, emails and calendars within Google Workspace. They will also bring Zapier-like automations native to Workspace. Google’s Notebook LM is bringing AI agents to help automate tasks as well.
It is hard to keep up with all the agentic AI tools being released. MIT has published an AI Agent Index at to help document technical components, intended users and safety features of agentic AI systems. The database is only updated to December 31, 2024, but it helps show the many agentic tools in development. The index framework helps assess agentic AI tools, including guardrails and oversight.
As we look to the future of Agentic AI in legal practice, the potential developments are both exciting and transformative. The rise of agentic AI, as demonstrated by Google's Notebook LM, promises to expand software capabilities and redefine how legal professionals interact with technology. At Legalweek 2025, discussions highlighted the embedded and agentic generative AI tools that are set to revolutionize the industry. Microsoft's Copilot is also paving the way for AI agents that can perform tasks autonomously, offering a glimpse into a future where AI seamlessly integrates into daily workflows. The National Law Forum emphasizes that the next generation of AI agents will bring unprecedented efficiency and innovation to law firms. As these technologies continue to evolve, legal professionals must stay informed and adapt to harness the full potential of agentic AI.
Agentic AI signifies a notable change in legal technology. It involves more than just drafting documents more efficiently; it entails changing the approach to legal work. These systems offer potential for innovation, cost reduction and improvements in client service, but also pose risks regarding privacy, ethics and professional responsibility.
Lawyers are not required to become technologists, but they need to be knowledgeable about how agentic tools are used. This involves addressing questions related to transparency, accountability and the balance between autonomy and supervision. The effectiveness of agentic AI in law will rely not only on the capabilities of these tools but also on responsible usage.
