Tuesday Riffs on Vertical Tech #1 - Legal AI, Voice Agents, and GovTech
riffs and news
Trying something new out. Let me know how it goes. Questions with some riffs. Riffs and some questions. You get the point. While we mainly cover strategy, companies, and industries in anchor pieces, the intent here is to riff on topics that I am still pondering or maybe don’t make up a full piece. Let’s consider the questions that everyone is thinking but won’t feature in a market map or press release.
Legal AI
Harvey is reportedly at 50m in revenue. If you read the relevant subreddits (yes, this is usually the best way to find pricing information), Harvey is roughly priced at $1,200 per user per month. Doing some simple math, we end up with roughly 3,472 users on Harvey. Doing some more basic research, one of their largest partners is Orick - comprising 1,000 lawyers with whom Harvey seems to have some sort of agreement to white label the platform. 1 big law firm and potentially a smattering of a dozen or so other clients.
There’s two ways to read this from an investor perspective:1
This seems like you have found a way to go sell mega deals but I’m not convinced you have found true PMF. What is the actual utility? If the opportunity in legal is so big, why are we already pursuing professional services users?
They’re already doing $50m on 3,472 users. We should give them a $3B valuation and another 300m dollars to go continue to sell megadeals. Their team is incredibly sharp and they will continue to make the product best-in-class.
I think you can see quickly which route was pursued. And for the record, I have no idea what to think. I am not a skeptic nor am I a true believer yet. It is incredibly obvious that if you assume a) the price point scales, b) the product is actually that useful or will be that useful, and c) they can continue to sell megadeals, why you would invest immediately.
On the one hand, Harvey’s price point seems high. On the other, it doesn’t seem high enough. A $1,200 per month per user arrangement for a BigLaw firm works out to breakeven productivity increases of ~6%. That seems achievable, but I don’t know if that is the case today. On the other hand, there’s theoretically no limit to productivity gains given model progress. That 6% could easily become 20%. In which case, Harvey should probably charge 2x their current price.
The point of this riff (can riffs ever have a true point?): in every single vertical AI company, it’s 2000 again. It’s quite difficult to tell at this point solely from revenue figures, the true defensibility, product success, or much about the future productivity achievements that we can expect.2
There will be true believers and true skeptics. The truth is somewhere in the middle. But frankly, Harvey feels like the sort of company that is good to go long on anyways.
Voice Agents
How you feel about voice agents is directly correlated to how your last phone call went with a customer service department. If you wanted to throw your phone against the wall, hearing the soothing voice of Scarlett Johanssen (what could have been…), sounds like a dream.
So many vertical voice agents are cranking and much like Harvey, I have no idea what to think about them. Yes, this beats an offshore team that I have no desire to talk to. No, I’m also not convinced this is the future… or rather, this current instantiation probably isn’t where we end up.
The problem is pretty simple: much like email, the more people who use voice agents, the less they work as a successful channel. This is most true on the sales side of voice agents. Good luck everyone. I wish you the best at getting me to answer an unknown number that already gets 10 spam calls a day. The worse this gets, the lower ROI voice agents theoretically become.
Where this potentially does work is if and when, voice agents are two-sided. If Apple ever releases a phone-answering agent and if my phone agent can call your phone agent, or if my phone agent calls your phone agent who calls another phone agent, maybe this works. Or probably, much like email, it ends up becoming a channel where 2% of conversations actually matter and triaging the channel becomes most important.
American brands punted on customer service being a branded value add to their customers long ago. It’s not surprising to see them adopt a new channel that does beat their offshore teams. I am surprised that so many SMBs are adopting these. Your brand is your only arbitrage as an SMB. I frankly, don’t think the voice tech is fully there to do this yet in a branded fashion. This may say more about the state of SMBs then it does the value of voice agents. Margins may be so tight, that companies can’t help but do this. But again, it’s a riff. You can’t hold me responsible for being wrong.
Local DOGE
Tyler Technologies continues to rip. I wonder what Elon would think of them. Apparently Tyler is wondering that too. From their latest earnings:
While the new Federal Department of Government Efficiency or DOGE, is still in its early days, we don't envision its effort having a significant impact on funding or demand for our software and services, which power generally essential functions.
This seems like the sort of thing you say if you know that your software isn’t exactly promoting government efficiency. “There will be no impact to shareholders, but also we know why you’re wondering!” But again, just riffs.
And perhaps, many local government leaders are beginning to think about this too. If you took an engineer who had spent his summer dissecting the Vesuvius scrolls into every municipality, what would their conclusion be about the current state of municipality software? And then there’s the political dynamics: You know how good your campaign pitch is if you say you are DOGE-ifying your suburb? People would look at you as a hero.
For these basic reasons, you better believe this is a question that’s going to feature heavily into new govtech upstart’s sales pitch. The feds have Elon. You have us. Let’s DOGE-ify San Francisco. Our software is the best way to do it.
And you know what? I hope it works. AI-enhanced government services with much better software UX is so obviously good that I wish these companies nothing but success. Go DOGE-ify every municipality. Go pitch hard against Tyler Tech. And heck, go partner with Stripe whom I’m sure would love nothing more than to power governmental payments.
Software can be so much more efficient, more cost-effective, and improve the quality of life for citizens. May your local DOGE efforts flourish, startups.
The resident investor readers please feel free to share alternate readings.
This is of course due to how many firms are going to have to rewire their internal dynamics to take advantage of AI. In no way is this Harvey’s fault but it does mean they will need to help firms achieve that sociotechnical revolution.