ILTACON 2026 went down in Nashville with one question at the center of it all: when AI acts on its own, who’s actually leading the flock?
The conference brought together legal industry experts to tackle a topic that goes way beyond chatbots and automated responses. The event, a global benchmark for legal technology professionals, surfaced discussions that blend technical innovation with practical accountability. And it became crystal clear that the market is still in its early stages of understanding what it truly means to put an AI agent to work autonomously inside a law firm.
The difference between generative AI and agentic AI might sound like just a technical detail, but mixing them up in practice is a mistake that can cost any firm dearly. Generative AI responds. Agentic AI acts. And that distinction changes everything.
That was exactly what Valerie McConnell, Vice President of Solution Engineering at Thomson Reuters, highlighted during the event. According to her, a few years ago the state of the art was asking a chatbot a question and getting an answer back. Now, we’re talking about AI systems that can take an objective, break it into smaller pieces, and arrive at an answer. We’ve moved from an AI that responds to an AI that acts, and that shift raises the stakes on technology adoption considerably. 🤔
And that’s precisely why the panel Secret Agentic Man: When AI Starts Acting on Its Own became one of the most relevant moments at this year’s ILTACON. The conversation got straight to the point: how much human oversight is needed for agentic automation systems to be truly reliable inside the legal environment?
The answer, of course, depends on how the agentic system is being used. What did become evident, though, is that the ideal solution combines lawyers’ legal expertise and client relationship management with the ability to automate lower-complexity tasks.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
What is Agentic AI and why it changes the rules of the game
When most people think about artificial intelligence in the workplace, they still picture that assistant answering questions, suggesting text, or organizing information on demand. Agentic AI is a leap beyond that. Instead of waiting for a command at every step, it receives an objective — like reviewing a contract, identifying risk clauses, and generating a report — and breaks that objective into subtasks it executes in sequence, without needing to be guided at each stage.
Leyla Samiee, Chief Product and Technology Officer at software company Introhive, laid out how this transforms the legal workflow. In the past, the lawyer defined the objective, found the path, and executed. With today’s agentic AI, a human defines the objective, but the model finds the path, and the execution trigger is decided by the lawyer in advance. And with the autonomous AI of the future, even that ideal trigger could be decided by the AI itself.
She put this in practical terms with a really interesting example. Imagine a partner leaves a firm after eleven years. Clients are impacted by a shift in work style and communication, but the newly assigned attorneys only realize it when it’s already too late. Now imagine an agentic world where an agent has access to call and email patterns to identify whether the tone of discussions is changing. Or even on day one, when the agent detects a user ID being deactivated and can pinpoint where the firm might be at risk.
As Samiee put it, the agent didn’t make the decision, but it provided enough information for the human to decide. That’s the key point of the technology. 🎯
The productivity gains are real and significant. But so is the risk. Because when an autonomous system makes a wrong call without anyone noticing in time, the damage can be substantial before any correction is possible. And in law, where every single word in a document can have concrete and lasting consequences, that risk needs to be managed very carefully.
Setting boundaries before unleashing the agents
All of this, of course, requires a lot of trust. Ali Mohammed, CEO of software company JuristAI, explained that when it comes to governance, he tries to balance risk and cost. According to him, when you delegate or hand off that kind of responsibility to something that will operate on its own, the agent is going to need to touch many different surfaces to develop a competent knowledge base.
This means that firms considering agentic AI adoption need to be proactive in determining which data the agent can and cannot access. What if it mixes information from different clients? What about data export control rules? What if certain clients are opposed to AI use? As Mohammed highlighted, this kind of thing needs to be defined deterministically before turning the agent loose.
Building those guardrails can be easier in theory than in practice. Some pioneering firms that ventured into this space have already run into unexpected ethical dilemmas. Mohammed added an important reflection: as we move into this new world where agents look more like coworkers than tools, we need to remember that while they may be colleagues, they are not human. And it will increasingly become part of a lawyer’s professional responsibility to teach these agents what they can and cannot do.
Reversibility and the one-way or two-way door test
With AI continuing to advance, this kind of planning may be just the beginning. Victoria Grech, Chief AI Officer at software provider iCompli by LegalRM, was direct with attendees when she said the future of work looks nothing like what we can imagine right now. She expects mixed teams of agents and humans to be part of the routine within a few years, with agent accounts already being developed for major platforms like Microsoft, Salesforce, and others.
However, there is a risk with these agent-centered teams. If swarms of agents decide there is no leader, they become even harder to govern. Grech used a metaphor that gives the debate its name: if there’s a flock of birds flying, who’s leading that flock?
She also pointed to a central technical challenge: while we’re building agents today, we can still see where they went wrong. But at scale, interpretability becomes a problem. And she left a question hanging in the air that resonated with attendees: at the pace we’re moving with autonomous agents, are we going to get ourselves into trouble?
To counter that risk, Grech championed the importance of reversibility — meaning the ability to correct a mistake that agentic AI might make. She thinks about it as a one-way or two-way door: any task a lawyer delegates to an agent should be a two-way door, one that allows the professional to walk it back if needed.
A practical example helps make this clear. An agent can be tasked with summarizing client emails and developing an action plan. That’s a two-way door, because the lawyer can reverse the summary or the plan if something is off. But if the agent is also tasked with replying to client emails without human review, that becomes a one-way door — and that’s not easy to undo after the fact. 🔍
Ethics and Governance as pillars of responsible automation
If there was one theme that ran through virtually every session at ILTACON 2026, it was ethics and governance in AI use. And not in an abstract or philosophical way — the discussions were practical, oriented toward what firms and legal departments need to do right now to build a solid foundation for responsible use of the technology.
The panelists agreed that now is the time to start putting policies in place that help govern the parameters of these agentic teams. McConnell, from Thomson Reuters, outlined some practical steps for firms looking to get into this space:
- Define permitted use cases, always ensuring action reversibility;
- Require observability — meaning use agentic systems with audit trails and decision tracking;
- Create accountability mechanisms, especially making sure lawyers cannot spin up new agents on their own without the firm’s involvement.
There is a balance between risk and reward, as the panel highlighted. McConnell was honest in acknowledging that we’re at a stage where it’s not possible to absolutely control everything the agent does. And she followed up with a spot-on line of reasoning: it doesn’t make sense to double-check every single thing the agent does, because then what would be the point of having a more efficient system?
At the same time, the AI agent integrated into the team should ultimately serve the lawyer’s needs. As Mohammed from JuristAI concluded, lawyers are craftspeople — true artists who have spent years developing their own style of practicing the profession. And precisely because of that, human oversight is going to become more important than it has ever been. ⚖️
ILTACON 2026 made it clear that the legal industry is at a real inflection point. Agentic automation is not science fiction or a project for five years down the road — it’s already available, already being tested, and already producing concrete results at firms that chose to move forward with care and strategy. The challenge now isn’t technological — it’s organizational and cultural. It’s about understanding that adopting agentic AI responsibly requires investment in governance, team training, process redesign, and a mindset that positions human oversight not as an obstacle, but as the element that makes all of this sustainable in the long run. 🚀
