Supposedly, the word of the year in technology is “agentic.” By the end of the year it’ll probably be “bubble,” but for now the world is meant to tremble before the awesome promise of agentic. It takes AI to a whole new level! It heralds the new human-free workforce! It keeps Woody Harrelson out of the rain! As the shiny object economy grows complacent with magic interactive chatbots, the agentic AI era to transform those humdrum chatbots into something… else.
The hype machine gods must feed and are only satisfied with the blood of freshly squeezed buzzwords. What does “agentic” even mean? “The term ‘agent’ is one of the most egregious acts of fraud I’ve seen in my entire career writing about this crap, and that includes the metaverse,” writes tech journalist Ed Zitron and, somehow, he might be too forgiving. The agentic talk means everything and nothing all at once. That’s the power of an empty signifier! Heap whatever vague hopes and dreams you can into this rhetorical nugget and let it carry you to a sale and another round of funding.
In theory, agentic AI represents the leap from autocomplete-on-steroids to autonomous action — taking an understanding of the user’s goals and setting out on its own to get the job done. Imagine a cybersecurity bot surveying the evolving threat landscape and inventing its own countermeasures as new viruses emerge all while you’re still fumbling with two-factor authentication. Other tech companies promote agents that monitor your calendar and autonomously decide to book your dinner, or buy flowers, or plot the grisly demise of your enemies.
But since lawyers entrusting this sort of autonomy to AI would be, to use the technical term, “malpractice,” agentic AI doesn’t really sing to this market. If you think hallucinated case citations are bad, wait until the agent takes it upon itself to settle your client’s divorce for pennies on the dollar based on Kirshner. Even if you’re not in Kentucky.
So why would anyone hoping to sell to lawyers — which presumably describes the folks on the ILTACON exhibit floor — indulge in the term “agentic” at all?
For the most part, it’s because their agentic AI isn’t really agentic AI. At least not the way anyone trying to fuel a half trillion dollar valuation would use the term.
We first encountered this a couple months ago when we saw Thomson Reuters preview its “agentic” offerings. Despite adopting the word-of-the-moment, the Thomson Reuters product felt more like a glorified automation product — which is good! Feed it case files, it does some research, it brainstorms some claims or defenses, does a little more research, and spits out a draft. That’s useful and can jumpstart a lawyer’s work, but it’s not so much an autonomous agent. If one were so inclined, they could enter “based on these case files, come up with claims, perform relevant legal research, and present me with a first draft motion” into a bot and (provided it had some way of accessing valid legal research other than its own fever dreams) get the same result right now.
That’s not so much agentic as it’s a batch file.
At ILTACON, Lexis showed off a similar feature and added a bit more to the agentic soup by explaining that they use different models for different tasks, making the request more complicated than simply asking ChatGPT a multipart prompt. Fair, but that’s still not an “agent” as much as professionally designed prompting.
When I raised my irritation with the whole agentic conversation with Tiana Van Dyk, Senior Director of Client Services at Epiq, she characterized (at least within the legal context) agentic as automation with prompts. “Is it that simple?” she mused. “There’s more in there, because if you have all the wrong models tied to all the wrong steps, you’re gonna get a bad outcome.” From her perspective, the challenge in legal tech right now is demystifying these technologies for lawyers. “What we’re missing in the industry, is our ability to translate the academia and the complex nature of some of these systems into something that is accessible.”
Though with agents the battle isn’t with the academics as much as the advertising execs. They’re the ones pitching autonomous decision makers to the world and the legal tech vendors have to come back and tell the lawyers, “ha ha, yeah, no, our product really doesn’t do all that.”
It’s frustrating, because it sounds like I’m downplaying the significant talent and expertise that goes into making these processes work right. Automation is hard! Making sure the process gets it correct every time takes a lot of strategic thinking and meticulous effort. But I keep reminding myself that it’s not my fault that this sounds dismissive, it’s the wholly unnecessary decision to set the bar at HAL 9000.
That’s not to say there aren’t niches within the legal workflow that could use genuine agentic AI. Cybersecurity is still a legal industry concern. The billing process might be ripe for something at least closer to agentic. Oddr doesn’t lean on the word agentic in their materials — hurray! — but its end-to-end revenue platform that automates the firm through billing, collections, payments and forecasting is much closer to the level of streamlined automation through several different systems than what many companies would plaster the word agent all over. Still, it’s not “evolving” its approach to the billing process unless it independently figures out how to call Rocco and Vinnie to pay a visit to that one delinquent client.
Legal tech vendors have a Scylla and Charybdis problem. On one side, they’re tech companies who feel compelled to go with the tech trend flow or look like they’re not “cutting edge” enough. On the other side, they cater to a legal profession that needs to be coaxed into tech adoption like frightened bunnies, soothed by reassurances that “this won’t really change how you’ve done this job for the last hundred years.”
The good news, is that lawyers can relax. Healthy skepticism is always welcome, but don’t let the “agentic” branding — and all the tech industry baggage that comes with it — scare you off of legal tech products. The providers who specialize in this space didn’t suddenly forget who they’re dealing with. They know that firms aren’t in the market for a product that’s going to start trying to “help” by making legal decisions on its own. Like Old Man Jenkins, the carnival caretaker who would’ve gotten away with it if it weren’t for you pesky kids, if you look underneath the hood of that “agentic” offering and you’ll find familiar prompts driving a familiar, vetted automation.
Useful, maybe even transformative at the margins, but not something making decisions.
The real danger is that buzzwords like “agentic” — and the delirious bucket of magic beans promises that comes with it — will scare lawyers into missing out on useful tools. But it’s the nature of this industry that we have to constantly push back against promises that thrill venture capital while terrifying lawyers. By next year there will be some new empty signifier. Symphonic AI… turning multiple agents into ebbing and flowing actors within a unified, goal-oriented symphony of voices. Or some equally bullshit phrase.
That is, unless “bubble” really does take over first.
Earlier: Dispatches From The AI Bubble: ILTACON 2025
‘Agentic’ AI Is The Hot Buzzword… But Do Lawyers Actually Want An Agent?
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news. Joe also serves as a Managing Director at RPN Executive Search.
The post Agentic AI Is The ‘Fetch’ Of Legal Tech And We Need To Stop Trying To Make It Happen appeared first on Above the Law.