The Rise of Agentic Systems and the Strategic Redesign of Legal Work
Introduction
In Part 1 of this series, we traced the evolution of legal technology from its mid-20th-century origins through the late 20th century, highlighting how early innovations like word processors, fax machines, and computer-assisted legal research revolutionized law practice, but only to a point. These tools were primarily adopted to pursue efficiency, speeding up manual tasks and eliminating obvious bottlenecks. This laser-focused on efficiency, however, proved to be an efficiency trap: it digitized existing workflows without fundamentally improving them, and often ran up against cultural and economic barriers, such as the billable hour, that discouraged time-saving innovations. Part 1 demonstrated that while these early technologies enabled faster paperwork and research, they rarely translated into a strategic advantage for firms, as the underlying service model, billing by the hour for tasks, remained unchanged.
Part 2 explored the modern era of workflow automation and how a more strategic approach can unlock value beyond simple speed-ups. We saw that automating multi-step legal processes (from client intake and document assembly to case management and billing) can yield not only time savings but improved accuracy, consistency, and client satisfaction when implemented with clear goals. For instance, strategically deployed automation reduces human errors in documents, accelerates client onboarding, and frees lawyers from routine tasks, allowing them to focus on higher-value advisory work. The key lesson from Part 2 was that efficiency should be treated as a byproduct of innovation, not the sole end goal. By aligning technology initiatives with firm strategy – improving quality, client experience, risk management, and data-driven insights – legal teams can achieve competitive advantages that far outweigh raw time saved.
Now, in Part 3, we turn to the cutting edge: agentic systems. These are AI-driven “agents” capable of operating with a degree of autonomy and proactivity in legal workflows, heralding what many call the next evolutionary leap in legal tech. In this article, we will examine how agentic systems differ from traditional automation, how they enable autonomy in legal workflows and collaborative multi-agent processes, and how their rise is driving the development of new service models, such as AI-native law firms. We will discuss the profound impacts on legal talent development and outline strategic responses for traditional firms facing this emerging paradigm. In doing so, we’ll integrate real-world examples of agentic innovation – from a fully AI-driven law firm automating small claims to large firms deploying AI “colleagues,” to illustrate the transformative potential and challenges of agentic workflows in law.
Autonomy in Legal Workflows: From Automation to Agentic Systems
Traditional legal automation has been largely deterministic: software executes predefined steps or responds to human prompts in a narrow scope. By contrast, agentic systems introduce autonomy, the capacity to make context-driven decisions, adapt, and act proactively within set boundaries. Often dubbed the “third wave of AI” in law, agentic AI involves deploying one or more AI agents (autonomous software entities) to manage a sequence of tasks towards a goal, with minimal micro-management.
In practical terms, this means moving from tools that simply assist humans to systems that can independently carry out objectives under specified conditions.
An agentic legal workflow might, for example, autonomously handle the drafting, filing, and management of a routine motion in litigation: the agent can research case law, generate a draft motion, file it via e-filing systems, diarize deadlines, and monitor the docket. All without a person directing each step. This is a qualitative shift from earlier automation (which might auto-fill a template but never initiate action on its own). Agentic AI can plan multi-step tasks, learn from feedback (utilizing techniques such as reinforcement learning), and adapt to changes in context in real-time. Crucially, these agents operate within defined parameters set by their human designers; they have autonomy, but it’s a bounded autonomy with ethical and practical guardrails. For instance, an AI agent may be empowered to draft and send certain client communications on its own, but only within a template and policy framework approved by the firm, and it might be required to obtain human sign-off for anything unusual (admittedly this is still in its nascent stage and there are risks to be mindful of as users and designers).
The autonomy of agentic systems enables them to be proactive problem-solvers in legal work. Rather than waiting for a lawyer’s instruction, an AI agent might continuously monitor relevant information streams and take initiative. It could detect, for example, a new court decision pertinent to an active case and immediately flag it to the team, along with an analysis of its impact. Or it might notice approaching contract renewal dates across a portfolio and automatically begin drafting renewal documents or compiling risk reports. By initiating action, these agents help lawyers transition from a reactive posture to a more proactive and preventive approach to the practice of law. This autonomy does not eliminate the need for human lawyers, but it does redistribute labor: the AI handles the heavy lifting of process and information, while humans handle supervision, nuanced judgment, and final decisions.
It’s important to note that today’s agentic AI, while advanced, is not a sci-fi superintelligence operating unchecked. These systems remain tools that augment human capabilities rather than replace them outright. They require careful design to ensure they stay within their lanes. Clear escalation protocols must be in place (e.g. if an agent encounters an unexpected scenario or a high-risk decision, it should pause and request human input). As we’ll discuss later, building trust in such autonomous systems is a major challenge, one that hinges on transparency, reliability, and maintaining a strong “human in the loop” for oversight. Nonetheless, the ability of AI agents to accomplish complex tasks with minimal supervision marks a significant turning point. It lays the foundation for the strategic redesign of legal workflows, where processes can be re-engineered around what technology does best (data processing, pattern recognition, and routine decision-making) and what humans do best (critical thinking, advocacy, ethical reasoning, and client relationship management).
Collaborative Agents and Multi-Agent Workflows
Agentic systems do not operate in isolation; in fact, one of their most powerful aspects is their ability to collaborate with humans and with each other. We envision hybrid teams comprising human professionals and AI agents working in tandem. For example, an AI research agent might scour databases and feed the most relevant findings to a human lawyer or to another AI agent tasked with drafting. That drafting agent could then produce a first-cut contract or brief, which a human lawyer reviews and fine-tunes. Such multi-agent workflows may involve a sequence of specialized AI agents, each contributing its expertise (research, analysis, drafting, project management), passing tasks among themselves much as human team members would. Human oversight remains the anchor of these workflows – lawyers set the objectives, define constraints, and handle exceptions or final approvals.
This collaborative paradigm extends to scenarios where agents coordinate with one another at machine speed. In an end-to-end contract negotiation managed by agents, one AI agent could handle initial drafting based on predefined playbook clauses, another could review incoming redlines from a counterparty’s mark-up, and a third could suggest compromises or identify fallback provisions, all while a human lawyer monitors the high-level progress. If the negotiation reaches a sensitive point (e.g., a non-standard clause with significant liability implications), the system may alert a human attorney to step in. Here, the agents function almost like junior colleagues or paralegals, collaborating to move the matter forward efficiently but deferring to human judgment for the tough calls.
Collaboration also involves clients and other stakeholders. Some agentic workflows directly interface with clients or end-users. For instance, an AI-driven legal assistant might interact with a client via a chatbot, gathering information or answering basic questions, then loop in a lawyer for more complex advice. In Garfield.Law’s case (discussed further below) illustrates that each automated action in a small claims matter still requires client approval, effectively making the client an active participant in the agent-driven process. This kind of structured human-agent collaboration ensures transparency and preserves trust: the agent does the legwork, but the human (lawyer or client) remains informed and in control at key decision points.
It’s worth noting that multi-agent architectures, where several AI agents with distinct roles collaborate, are becoming increasingly feasible with advances in AI. Large tech providers are starting to offer platforms for orchestrating multiple AI agents on legal tasks. For example, KPMG’s in the US is leveraging Google’s AI to scale multi-agent platforms for tasks like contract review and compliance checks. The appeal of multiple agents is specialization: rather than one monolithic AI trying to do everything, you have a team of narrow AIs, each excellent at its niche (one for legal research, one for drafting, one for project management, etc.), communicating with each other. This mirrors the traditional law firm model, where different specialists collaborate on a case, but with the added efficiency of machines. The coordination among agents (and with humans) is orchestrated according to a workflow design, which itself becomes a new kind of legal practice competency: knowing how to architect and manage a human+AI workforce.
In sum, agentic workflows thrive on collaboration. They are not about AI acting alone in a vacuum; they are about AI acting together with other AI and with people to achieve outcomes more effectively. This demands not only technical integration of systems but cultural integration within firms: lawyers must learn to trust their digital assistants (up to a point) and to supervise them intelligently. At the same time, technologists must tailor AI agents to fit the realities and nuances of legal work.
AI-Native Firms and New Service Models
One of the most fascinating developments in this agentic era is the rise of “AI-native” law firms, built around agentic systems, is reshaping legal service delivery. These firms are not merely using legal tech as support; rather, AI is central to their service delivery model.
- Garfield.Law: AI-Only Small-Claims Practice: Authorized in England & Wales in May 2025, Garfield.Law automates every step of sub-£10k debt claims, from demand letters to court filings. Software agents act as autonomous practitioners, with a supervising solicitor and client approvals providing guardrails. The result is a scalable, low-cost service that broadens access to justice. This model targets high-volume, low-value cases that traditional firms often find unprofitable, thereby expanding access to justice in an area of unmet need. Dr. Corsino San Miguel, commenting on Garfield’s approval, described it as a “redefinition of what it means to be a law firm in the age of AI,” noting how it challenges the profession to rethink service delivery in a digital landscape.
- Covenant: AI for Private-Market Investors: Launched in 2024, US-based Covenant pairs multiple AI models with expert lawyers to review LPAs and related contracts. Agents extract key terms, flag risks, and draft markups; humans verify and advise. A two-day turnaround, SOC 2 certified security, and lower fees have attracted endowments, foundations, and sovereign wealth funds. Covenant shows how AI-centric managed services can outpace traditional firms. Covenant’s success points to an emerging service model where legal work is productized into AI-enhanced tools and platforms, offered at a lower cost and at a faster pace, disrupting the traditional, high-margin, time-intensive approach of elite law firms.
- Avantia’s “Ava”: In-House Agent at a Global Firm: Asset management specialist Avantia unveiled “Ava” in January 2025. Trained on nearly one million internal documents and embedded in Microsoft 365, Ava drafts emails, checks compliance, and suggests precedent aligned edits. Senior lawyers review every output, freeing them for higher-value work and speeding up deals. Avantia demonstrates how incumbents can become AI-native from within by building proprietary agentic technology. Avantia’s approach demonstrates how a traditional firm can evolve its service model by developing proprietary agentic technology: rather than relying solely on off-the-shelf legal tech, they built a bespoke AI tuned to their practice. This provides a competitive differentiator (expertise + custom tech), and it highlights that incumbent firms can proactively adopt an AI-native mindset. In effect, Avantia is becoming an AI-native firm from within, rather than being born as one.
Taken together, these examples tell a simple story. Agents deliver best when their playground is narrow. Yet volume and velocity are only half the equation. Pairing automation with seasoned lawyers preserves quality, safeguards client relationships, and brings the nuanced judgment that software cannot yet replicate.
That blend is already reshaping the business model. Fixed-fee and value-based arrangements are edging out the billable hour, aligning costs with outcomes that AI-driven teams can predict and achieve. Clients may soon judge firms against AI-assisted alternatives that promise faster turnarounds and lower prices, prompting traditional practices to either justify a premium for human-only work or evolve with their agentic technology.
Implications for Legal Talent and Skill Development
Agentic AI is rearranging the talent map of law. As software agents take over routine research and first‑draft work, lawyers step higher, focusing on oversight, strategy, negotiation and client counsel. Freed from drudge tasks, junior associates can observe real legal reasoning sooner, while partners spend more time on complex judgment calls that clients value most.
This shift brings fresh opportunities. Automation reduces the need for document review or case updates by 80 percent or more, easing burnout and unlocking capacity for higher-order analysis. Data-rich workflows also invite new hybrid skills: reading analytics, refining processes, and managing AI tools now sit beside traditional legal analysis. Lawyers who can critique an agent’s output, fix bias, and craft better prompts add a new kind of expertise to their resume.
Yet challenges remain. The old apprenticeship model relied on manual grind to build deep knowledge. Firms must redesign training, mixing simulated exercises, close mentoring, and varied rotations so juniors still learn the substance behind the software. Teaching “AI supervision” is essential: lawyers must question sources, spot hallucinations, and know where the tech fails. Change management looms large; adoption will be uneven, what Ethan Mollick calls a “jagged frontier,” where early wins coexist with pockets of skepticism and workflow friction. Leaders have to adjust incentives, refresh billing models, and provide hands-on coaching so new habits take root. Cultural resistance also lingers; prestige-minded practitioners may bridle at delegating to machines. By framing AI as an amplifier, not a threat, and involving younger lawyers in refining its output, firms can ensure human judgment grows alongside automation.
Technology alone won’t move the needle unless people, processes, and incentives evolve with it. Expect an uneven landscape, Ethan Mollick’s “jagged frontier,” where breakthrough wins sit next to stubborn habits and misaligned billing models. Effective leaders roll out agentic AI through bite-sized pilots that demonstrate value quickly, pair them with transparent communication, and reward the new behaviors they want to see. Internal champions, continuous feedback loops, and clear guardrails on risk help turn early experiments into sustainable practice.
Strategic Responses by Traditional Firms
AI‑native providers have moved the goalposts. To stay relevant, established firms must treat agentic technology as a board-level priority, not an IT upgrade. First, articulate a purpose that goes beyond shaving costs: ask how AI can improve client outcomes, open new service lines, or enhance lawyer wellbeing. Clear intent guides every downstream choice.
Second, learn by doing. Small pilots. For example, an internal knowledge‑bot or a fixed-fee NDA agent that generates real-world data on accuracy, workflow impact, and client appetite. Quick wins build internal momentum and surface process snags before wider rollout.
Third, decide whether to build, buy, or partner. Some global firms are hiring data scientists to craft bespoke agents tailored to their playbooks; others join forces with startups or Big Four affiliates that already have mature systems. The right mix depends on budget, scale, and appetite for experimentation; however, waiting on the sidelines is no longer a safe option.
Fourth, realign the business model. Automation collapses lawyer hours, so tying revenue to time becomes self‑defeating. Fixed-fee or value-based pricing allows firms to capture the upside of speed while providing clients with cost predictability. Transitioning requires fresh metrics and frank conversations with finance teams and clients alike.
Fifth, invest in culture. Lawyers respect precedent; innovation feels risky. Leadership must reward experimentation, share success stories, and train staff on both the “how” and the “why” of AI. Career paths that blend legal and tech skills keep talent in‑house. Solid governance frameworks, covering data security, bias monitoring, and accountability, reassure regulators and clients that the firm uses AI responsibly.
Conclusion: Beyond Efficiency to Strategic Value
Across this series, we traced the arc from digitization to automation to fully agentic AI. The lesson is simple: technology delivers lasting advantage only when paired with purposeful strategy and human judgment. Agents can draft, summarize, and predict at machine speed, but lawyers still supply context, creativity, and ethical reasoning. Firms that redesign workflows, pricing, and talent development around that partnership will unlock faster deals, richer insights, and happier clients.
The future will not arrive evenly. Expect jagged progress: breakthroughs in one practice area, skepticism in another. Managing this frontier calls for disciplined change management, transparent incentives, and continuous feedback loops. Those willing to pilot, learn, and iterate will shape best practice; those who ignore the shift risk watching their premium work migrate to more agile competitors.
Efficiency is a milestone, not the destination. The goal is to deliver strategic value, enabling better justice, sharper commercial results, and more fulfilling careers. By embracing agentic AI with clear intent and robust change leadership, traditional firms can turn disruption into a differentiator and write the next chapter of legal innovation.
