Last week Tom Martin, a seasoned legal technology expert with over 26 years of legal experience, shared insights about the emerging world of AI agents in legal practice. Martin, who teaches generative AI at Suffolk Law School and runs legal technology company LawDroid, provided a pragmatic perspective on how law firms should approach AI implementation. His discussion revealed both the transformative potential of AI agents and the strategic considerations necessary for successful adoption.

Starting with Why?

The Foundation of AI Implementation

Martin emphasized that the most critical question isn’t “how” to implement AI, but “why.” This foundational approach challenges the common tendency to chase exciting demonstrations or competitive pressures without clear purpose.

“The very first question is why,” Martin explained. “You need to ask why before AI. What’s the problem? What’s the friction point? What is the value proposition for the law firm or business that you’re gonna use AI to solve? And that really narrows the focus.”

This philosophy stems from observing firms that get “ahead of themselves” due to excitement and anxiety about AI’s rapid advancement. Rather than jumping on board simply to avoid missing out, Martin advocates for identifying specific organisational problems that AI can meaningfully address.

The practical application involves collaborative problem identification: “The work involved is actually creating… sitting down with everyone you work with and creating a list of all of the problem points, all the friction points, all the bottlenecks, all of the time sinks. And then amongst those, determining which one if solved, would provide the highest value.”

AI Agents vs. Generative AI: Understanding the Distinction

A significant portion of Martin’s discussion focused on clarifying what constitutes an AI agent versus traditional generative AI applications. This distinction is crucial as the legal profession enters what Martin calls “the year of the AI agent.”

Martin explained that many assume AI agents are simply workflows—where outputs from one AI prompt feed into subsequent prompts. However, true AI agents involve something more fundamental: discretion.

“An AI agent is where you actually give the AI model discretion,” Martin noted. “You allow the AI to actually employ its judgment in making a determination between a couple of options. Those couple of options could be qualifying a potential client, like do they fit your criteria or do they not fit your criteria? And you leave that discretion to the AI based upon the tools it has available to it.”

This discretionary capability represents a fundamental shift in AI application. While generative AI provides “help and assistance,” AI agents deliver “outcomes”—a distinction that explains much of the excitement surrounding agent technology.

The Implementation Strategy: Starting Small and Building Trust

“I believe that you should start small,” Martin emphasized. “Especially with the first project… you wanna experiment on it in a sandbox, hopefully, where there won’t be much harm caused if it goes south.”

This strategy extends to team involvement. Rather than relying solely on innovation officers or senior partners, Martin recommends engaging the entire organization: “You have to have everyone contribute because even someone who maybe is not highest on the totem pole… they can give you some really good ideas about how to save time or money.”

Interestingly, this democratized approach particularly benefits younger lawyers, who often possess greater familiarity with AI tools. Studies support this trend, with PricewaterhouseCoopers research showing that professionals with generative AI skills earn up to 52% more—creating new opportunities for junior lawyers in a market often concerned about AI displacing entry-level positions.

The Control Paradox

The concept of giving AI systems discretion presents a particular challenge for legal professionals, who Martin acknowledges are naturally control-oriented. However, he suggests that competitive pressures and fear of missing out (FOMO) are overcoming traditional caution.

“I think what’s overcome that for some of them is just FOMO,” Martin observed. “It’s radically strange right now, the environment we’re living in, where people are so concerned about competitive pressures.”

These pressures aren’t merely internal to the legal profession. External threats, such as the partnership between Perplexity and LegalZoom, demonstrate how technology companies are positioning to bypass traditional legal services entirely. As Martin noted: “People are already researching things through perplexity to get answers, and then to be hooked up with LegalZoom to actually provide the services. People may not even need lawyers.”

To address control concerns, Martin recommends building “breakpoints where there’s humans in the loop that can give their wisdom and experiences to how to take that next step.” This human-in-the-loop approach allows firms to leverage AI agent capabilities while maintaining oversight and intervention opportunities.

Regulatory Landscape: Navigating Uncertainty

The regulatory environment surrounding AI in legal practice varies significantly between jurisdictions, creating additional complexity for firms operating internationally. Martin expressed particular concern about the laissez-faire approach emerging in the United States.

“The approach right now legally is one of laissez-faire,” Martin explained. “With that new bill that may or may not pass in the US Congress, they wanna stop the individual states from regulating AI whatsoever for 10 years. And that kind of complete wild west approach is not something that we need right now.”

This contrasts with more structured approaches in other jurisdictions. The UK’s authorization of an “AI-first” law firm by the Solicitor’s Regulation Authority represents a different regulatory philosophy, though one that has generated significant debate within the profession.

The regulatory uncertainty extends to accountability frameworks. Traditional internet company exemptions may not apply to AI systems that actively generate content and make decisions. As Martin noted: “They actually did create the AI model. The AI model is saying particular things and taking particular actions, especially with AI agents. So they are assuming more of a product liability accountability.”

Client Expectations and Economic Pressures

Client attitudes toward AI adoption are driven primarily by economic rather than ethical considerations. While clients care about confidentiality and professional standards, their primary concern focuses on cost and efficiency.

“Their concerns are more about the money,” Martin explained. “When they see their friends getting some rental agreement just drafted up by AI immediately… their perception is that why am I paying for you? Why are you taking so many hours to do this?”

This client pressure coincides with significant fee increases across the legal profession. Post-COVID rate increases of approximately 40% at top-tier firms have intensified scrutiny of legal spend, particularly among in-house counsel managing external relationships.

The economic dynamics create a compelling case for AI adoption beyond mere competitive advantage. As Martin emphasized: “It doesn’t make sense for us to not make use of these tools to stay competitively… to stay at a competitive advantage.”

Strategic Implementation Areas

Martin’s practical recommendations focus heavily on non-core legal functions as initial implementation targets. These areas—including client onboarding, KYC compliance, CRM management, and phone answering—present lower-risk opportunities for automation whilst delivering immediate operational benefits.

“You could automate answering the phones, you could automate potential client qualification,” Martin suggested. “There’s a lot of that stuff that is busy work where you literally can’t scale to meet the volume of need that you might have.”

This approach allows firms to gain experience with AI systems whilst avoiding direct impact on core legal services. Success in these areas can build internal confidence and expertise before expanding to more sensitive applications.

The Technology Reality Check

Despite his expertise in AI development and implementation, Martin maintains a balanced perspective on technology selection. He emphasizes that AI isn’t always the optimal solution, advocating for traditional programming approaches when appropriate.

“Good old fashioned AI with conditional logic… is perfect for legal because your expectations are met every single time,” Martin explained. “One example of that is document automation. If you have a form template that needs to be filled out, you could use conditional logic for that, and it’s satisfactory every single time.”

This pragmatic approach acknowledges that generative AI’s probabilistic nature may introduce unwanted uncertainty in certain applications. The key lies in matching technology capabilities to specific use cases rather than pursuing AI implementation for its own sake.

The Transformation Imperative

While advocating for careful, problem-focused implementation, Martin also emphasized the fundamental nature of the current transformation. This isn’t merely another technology trend but a paradigm shift comparable to previous industry disruptions.

“This is not a fad, it is not hype. It is not something that’s gonna go away,” Martin stressed. “This is something that is a new way of doing business for everyone, including lawyers.”

Historical examples like Blockbuster and Barnes & Noble demonstrate how established industries can face displacement when new business models emerge. The legal profession’s traditional regulatory protections may not withstand the scale of technology companies entering the legal services space.

The scale of venture capital and private equity investment in legal technology underscores this transformation’s significance. The legal profession, long considered “the last bastion of old fashioned service providers,” has become “one of the hottest professional services being looked at by venture and PE.”

Educational Resources and Continuous Learning

Throughout the discussion, Martin emphasized the importance of education and continuous learning. He referenced Richard Susskind’s recent work “How to Think About AI” as providing accessible insights for legal professionals, while also mentioning his own textbook on generative AI and legal services delivery.

“I highly encourage everyone to get into those resources and learn more about this,” Martin advised. “Once you get to understand it a bit better, you can make much more informed decisions that aren’t motivated just by keeping up with everyone else.”

This educational approach supports more strategic decision-making, moving beyond reactive responses to competitive pressures toward thoughtful implementation based on clear understanding and specific organizational needs.

Conclusion

Martin’s insights reveal AI agents as representing a significant evolution in legal technology—moving from assistance tools to outcome-generating systems. However, successful implementation requires careful planning, clear problem identification, and strategic thinking about organisational needs.

Key takeaways for legal professionals include:

• Problem-First Approach: Identify specific friction points and value propositions before selecting technology solutions
• Start Small: Begin with low-risk, high-value projects to build experience and confidence
• Engage Everyone: Involve all organisational levels in problem identification and solution development
• Maintain Control: Implement human-in-the-loop systems to preserve oversight while leveraging AI capabilities
• Focus on Non-Core Initially: Target administrative and operational functions before core legal work
• Continuous Learning: Invest in education and understanding to make informed strategic decisions

The legal profession stands at a inflection point where AI adoption isn’t merely advantageous but necessary for competitive survival. However, the path forward requires balancing innovation with the profession’s fundamental responsibilities to clients and society. As Martin concluded: “Focus on concrete problems… not to get carried away with the excitement about AI and AI in general, but to focus on what we’ve always had to focus on: the problems that we have in front of us.”