Your AI and My AI, Sitting by the Fire. . .

Fractionally Legal 60: Reflections From a Week at the AI Shakedown Street, and Ticketmaster.

I spent the better part of last week in San Francisco visiting clients and attending Human[X], a huge AI conference where about 6,500 people spent 3 days fetishizing AI. I assume most participants have real jobs that either require them to be there to sell stuff or, like me, have real jobs that don’t (yet) involve thinking about AI all day and so were ducking in and out of the conference between actual work. However, as the conference demonstrated, the line between actual real work and AI is blurring. There is a good chance that in a few years you’re either going to have figured out a way to make AI work for you, or you’re going to be replaced by AI that does your work.

As I’ve written about many times before, I have a fundamental belief that AI is going to allow me to serve my clients faster and cheaper, so I am “all in” on deploying AI responsibly across my practice. Granted, I’m in a unique space as a fractional general counsel where I’m juggling lots of cash flow sensitive clients and they all want quality legal work, fast. In a nutshell, that is what I do. As I’ve written about before, if you have a multi-billion dollar deal where money is no object or you’re handling the trial of the century, AI is much less impactful. Clients are going to spend the money to throw human bodies at the problem. Same issue with the government: it has all the money and lots of humans to throw at any legal problem (I just paid my taxes, so I know about the money part!). If anything, for those clients, legal AI is actually going to make the work product worse, not better.

With that framing in mind, I went to Human[X] to hear from some of the brightest (or most enthusiastic) minds in AI. I have a few takeaways, especially for any lawyer who wonders what AI is going to do to your practice.

First, it’s important to note that there were a lot more sellers than buyers of AI products at the moment. That is the result of the frothy investment atmosphere. So if you’re looking for AI solutions for your law firm, you have the leverage. Side note: I bemoan the fact that we can’t seem to find the resources to build out physical environments (hence the high cost of healthcare and housing). But we certainly have ample resources to build new digital tools, most of which won’t go anywhere. That is why we can’t have nice things. We don’t invest in them.

Second is the current direction of the AI industry. The buzz in the real world is all about AGI: Artificial General Intelligence. But that is not what was being discussed at Human[X]. At Human[X] it was all about the “agents.” So let’s talk about the agents.

Agents are those little bits of code that can actually do things. AI powered agents are “agentic AI.” There are companies building “secure agents,” “fast agents,” “enterprise grade agents,” “efficient agents,” agents that monitor other agents. The product is essentially using AI to write code that actually accomplishes a task, and is smart enough to work autonomously — set the goal and it will figure out how to do it. AI can write code so fast that the agents can adapt and evolve, and then you need the agents to watch them. Apparently, building agents to watch agents is as important as building the agents themselves, which begs the question of how many agents are enough agents. Seems like the ultimate tautology and we really don’t know.

In the legal space, which I’m most interested in, there are companies that build agents to accomplish legal tasks — hundreds of thousands of them if the marketing is to be believed. Alas, the two big companies building agents, Harvey and Legora, were not actually at the AI conference. That’s probably because the market for specialized legal agents is smaller than one might think. The lawyers with the most money to buy agents probably need them the least because their clients will pay for human labor. Smaller practices like mine can really benefit, but the use cases are more limited than most people realize. I’m finding bottlenecks in my practice and building agents to solve them. Building the agents is not hard (if I can do it, anyone can). Identifying the use case is.

For example, with a few exceptions, I’d never build an agent that automatically responds to emails for me with AI-generated responses. Draft, yes (already have that) but I review everything before it goes out. Any time saved with an agent that sends emails autonomously would get eaten up trying to walk back a poorly worded email. Ditto with an agent that redlines and sends back a contract to a counterparty without my review. Most of the time, I’ll just send what the AI creates. But catching that one-in-a-hundred error will save me dozens of hours I’d otherwise spend trying to unwind the mistake. And remember, legal AI still has the fundamental problem of hallucinations. AI likes to make things up. I do my own research, make sure I have all the salient facts (AI helps me review documents to get the facts right), and turn the results over to AI to draft with specific instructions to only use the materials provided.

The most interesting AI I ran into at Human[X] was by https://samu.live which validates all the inputs to, basically, allows you to always go back to the source of any “fact” cited by the AI. Adopting some type of protocol like that would be a huge unlock for the use of AI in court proceedings. Wonky, but it’s that type of tech we need if we’re going to deploy legal AI at scale. Otherwise, we’re just building agents to prevent hallucinations, which seems like a big waste because we can’t really control the agents without building other agents. And so on. Think of autonomous agents as the rabid white whale of the legal AI industry.

And that leads to the third point: there is a lot of blind optimism and we’re still in the “gee whiz” phase of AI, especially AI for legal services. First, everyone assumes that there will be cheap LLMs and the energy infrastructure to support all these agents in the future. There was not a lot of recognition that there is a bubble, and all the experimentation is currently underwritten by generous investor dollars. Without that, most of this experimentation economy would just become too expensive to operate because developing and operating an LLM at scale is really, really expensive. Case in point: OpenClaw, an open source protocol that can use any LLM you want, even one stored locally on your laptop that you downloaded from China for free, was all the rage. It seemed like a tacit admission that once the price of running the LLMs goes up, the world changes.

The blind optimism is pervasive, and infectious. Very little was about the downside risk of turning too much over to agents. There was almost no talk of government regulation, as if the industry would regulate itself. The closest thing to a government official was Al Gore (yes, he is still alive and really vibrant) and even he was talking about whether we need to recognize AI consciousness as tantamount to human consciousness. It was interesting but, oh brother.

But the threat is real. Case in point: Anthropic just released a new model called Mythos that can identify and exploit zero-day vulnerabilities in every major operating system and web browser — fully autonomously, with no human involvement after the initial request. Anthropic’s response was to restrict access to roughly 50 vetted cybersecurity partners through a program called Project Glasswing, giving defenders a head start before broader release. No law required this. No regulator compelled it. Simply releasing Mythos into the world would not have been a crime — but it would have been negligent. Anthropic just decided it was the right thing to do. That’s admirable — but it’s not a governance model. Would Musk have made the same call? North Korea?

At Human[X], everyone assumed there’s some architecture that can keep agents in check. That’s a big assumption. This is not a safety-focused industry. And if this technology is as powerful as its promoters claim, we need serious regulation — now, not after the first catastrophic breach.

In the meantime, I’m going to keep doing what I do: use AI to serve my clients better, faster, and cheaper, without pretending the risks don’t exist.

If you want to see what that looks like in practice, check out www.lawyerloop.ai, where you can find independent lawyers use AI to deliver legal services at a price that actually makes sense. And in the coming weeks, I’ll be overhauling www.yourfractionalgc.com so you can better understand what a Fractional General Counsel can do for your business. I’m not trying to be an AI law firm. I’m trying to be a better lawyer. AI is how I get there.

Ticketmaster

Iran is bad, but the Ticketmaster verdict is a reminder that Trump and his sycophants are really bad at what they do. I’ve written about Ticketmaster a few times before — as an avid live music fan and repeated victim of their high fees and crappy technology I have a special ax to grind.

Even though the states won the trial and showed that Ticketmaster was an illegal monopoly and used that market power to harm consumers, we actually don’t know what the remedy will be. As is common in antitrust, there is a liability trial and a remedy trial. For reasons that probably have to do with a Palm Beach cocktail party and incompetence, the Trump DOJ essentially gave up and settled with Ticketmaster before the jury verdict, which means that they got a much worse deal than they would have gotten had they taken the case to verdict and gone through the remedy phase. The DOJ dropped its case in exchange for $280 million — paltry — and an agreement that the amphitheaters it owns can use other ticketing software and a sale of 13 of its venues. The states are pushing for a full breakup of the company.

The states are right. The real issue is the deals that lock in venues to use Ticketmaster, whether owned by Live Nation or not. Why would a venue not owned by Ticketmaster agree to use its crappy ticketing system and charge the high fees? Because if they don’t, acts managed by Live Nation, the country’s largest promoter, are going to venues that have deals with Ticketmaster or are owned by Live Nation. It’s all pretty simple. And because there is such a limited market for alternative software, no one builds one. Imagine spending the time and energy to build a better ticketing system when you have no customers. So Ticketmaster is the only game in town and charges exorbitant fees for a crappy ticketing product. It’s called market manipulation and it costs me and every consumer of live music lots of money in fees and lots of brain damage. I’m happy a jury understood that, even if Trump’s DOJ can’t, or won’t.

Keep thinking, keep building,

Jesse

Keep thinking, keep building,

Jesse

Hi, and welcome to my newsletter! I’m Jesse Strauss, Your Fractional General Counsel. I’m a lawyer with a private practice based in New York City, helping clients in the United States and globally with their U.S. legal needs. My expertise spans various areas, including raising funding rounds, employment issues, negotiating master service agreements, intellectual property, compliance, legal process management, and dispute resolution. My focus is on founding and nurturing great companies from seed to exit. Discover more at Your Fractional GC and book a complimentary 30-minute consultation. You can also follow me on Threads @lawyerjesse1977, on BlueSky @lawyerjesse.bsky.social, subscribe to my Substack here (follow me on notes), and follow me on LinkedIn here.