How AI is Transforming Labor Markets AI transcript and summary - episode of podcast a16z Podcast

Go to PodExtra AI's episode page (How AI is Transforming Labor Markets) to play and view complete AI-processed content: summary, mindmap, topics, takeaways, transcript, keywords and highlights.

Go to PodExtra AI's podcast page (a16z Podcast) to view the AI-processed content of all episodes of this podcast.

a16z Podcast episodes list: view full AI transcripts and summaries of this podcast on the blog

Episode: How AI is Transforming Labor Markets

Author: Andreessen Horowitz
Duration: 00:44:38

Episode Shownotes

Did you know the U.S. nurse labor market is over $600 billion annually, but the dedicated software market for nurses is almost zero?In this episode General Partners Alex Rampell, David Haber, and Angela Strange discuss how AI is revolutionizing labor by automating tasks traditionally done by humans.They’ll trace the evolution

of cloud eras — from the original to financial services-enabled to the current AI-enabled outcomes era — showcasing how AI is creating unprecedented opportunities, allowing startups to outpace incumbents. They also explore how this shift will reshape industries, where we are in the adoption curve and what companies need to succeed, and the gaps where the a16z Enterprise team would love to see more innovation.Resources:AI Turns Capital to Labor: https://a16z.com/ai-turns-capital-to-labor/The Messy Inbox Problem: https://a16z.com/the-messy-inbox-problem-ai-apps-wedge-strategies/Vertical SaaS: Now with AI Inside: https://a16z.com/vertical-saas-now-with-ai-inside/BarbAIrians at the Gate: The Financial Opportunity of AI: https://a16z.com/financial-opportunity-of-ai/Stay Updated:Let us know what you think: https://ratethispodcast.com/a16zFind a16z on Twitter: https://twitter.com/a16zFind a16z on LinkedIn: https://www.linkedin.com/company/a16zSubscribe on your favorite podcast app: https://a16z.simplecast.com/Follow our host: https://twitter.com/stephsmithioPlease note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures.a

Full Transcript

00:00:00 Speaker_01
Now you have software agents that are effectively doing what for 65 years have been human work.

00:00:08 Speaker_04
Is that going to increase software revenue 2x? It could potentially increase it 10x.

00:00:12 Speaker_00
It's not even on the same kind of playing field.

00:00:14 Speaker_01
All the data is here in the cloud. All the compute is in the cloud. And now you just mix them together.

00:00:20 Speaker_00
Many of these incumbents aren't going to evolve.

00:00:22 Speaker_04
Every time there's a new technology shift, we have to challenge every investment thesis that we thought was not going to work. And so many of them now are going to work.

00:00:31 Speaker_00
Moats still matter. And a lot of the moats in software today are the same that they've always been.

00:00:36 Speaker_01
It's both defense and offense for these companies to figure out what the hell to do.

00:00:43 Speaker_05
There's a lot of talk and investment in AI. But that opportunity is often framed in software terms, the budgets, the market caps, and the existing companies of the last era. But might this era be fundamentally different?

00:00:57 Speaker_05
Consider that for centuries, the biggest school of science was alchemy, to turn lead into gold. And today, AI enables a far more powerful transmutation, turning software into labor.

00:01:08 Speaker_05
Capital no longer just buys engineers or hardware, it buys code that replaces or augments labor, unlocking completely new markets.

00:01:17 Speaker_05
And that's exactly what we discussed in today's episode, together with three A16Z General Partners, Alex Rampel, Angela Strange, and David Haber.

00:01:26 Speaker_05
Together, we trace through the evolution of previous cloud eras and what those tell us about what's to come. Perhaps most importantly, what's really new here?

00:01:35 Speaker_05
Namely, that the $300 billion enterprise software market is just a fraction of the multi-trillion dollar white-collar labor market. These niche markets that weren't that interesting are now potentially very interesting.

00:01:48 Speaker_05
We'll also cover where we are in the adoption curve, wedges and defensibility for startups, and how pricing is being upended.

00:01:55 Speaker_01
If they don't do this right, they could lose all of their revenue, or most of it. If they do it really well, they could 10x their revenue.

00:02:01 Speaker_05
Finally, if you're building in this space, we'd love to hear from you. Just reach out to podpitches at a16z.com, and we'll route you to the right place. All right, let's get started.

00:02:14 Speaker_05
As a reminder, the content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any A16Z fund.

00:02:27 Speaker_05
Please note that A16Z and its affiliates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a16z.com slash disclosures.

00:02:44 Speaker_05
Alex, you wrote an article recently, Input Coffee, Output Code. But this idea of turning capital into labor, hasn't this always been true? What's new here?

00:02:54 Speaker_01
Well, it certainly has been true for a long time. If you watch some old movie about the Romans, you'd have all these Roman slave laborers or Roman soldiers rowing in unison on a boat.

00:03:05 Speaker_01
And then, of course, you got the steamship, and then you don't need these 50 people rowing anymore. So clearly, there has been this long historical arc of technology is, in some cases, augmenting labor.

00:03:18 Speaker_01
But it was always the brawn and not the brains, right? It's, I have a bunch of people that are sewing clothes and now I have the loom. But everything that was what we would now call white collar, that hasn't happened before.

00:03:29 Speaker_01
And what I talked about was the three or four different areas of software, just storing information. So for the longest time, if I wanted to keep track of who was on my airplane, from the Wright brothers onward.

00:03:41 Speaker_01
I would have a filing cabinet with, it's like, oh, here's Pan Am flight 192, and here's who's on it, and I'm writing down their name, and then they might call or send a telegram back in the old days of, I don't want to be on that plane anymore, then I erase it, and then I refile it.

00:03:55 Speaker_01
And one of the first examples of digitization of the filing cabinet was something called Sabre. And this was developed by American Airlines, I think in 1959 or 1960 in concert with IBM.

00:04:09 Speaker_01
And this was one of the first examples of taking this filing cabinet that really kept track of who's on Pan Am flight or American Airlines flight, in this case, and putting it in a database.

00:04:18 Speaker_01
So now, instead of having filing cabinets with lots of erasers and whiteout, you replace the filing cabinet with a computer.

00:04:25 Speaker_01
And this in turn begot travel agents and travel agencies because they had a thin client, a little computer, a terminal that I remember like booking airline tickets with my mom in the 1980s.

00:04:36 Speaker_01
You'd go into the travel agent and they had like a green screen computer that connected to a mainframe in Texas, which is where Sabre was. So the first realm of software.

00:04:46 Speaker_01
was really called from 1960 onwards, when computers became a thing, of take a filing cabinet, so it could be the HR filing cabinet, it could be the medical filing cabinet, it could be the financial filing cabinet, and put that in software.

00:05:01 Speaker_01
And what was that? It's a database with a front end to actually go enter things. So Quicken famously did this in the 1980s for financial statements. There was a company called PeopleSoft that famously did this.

00:05:12 Speaker_01
It was the first HR filing cabinet put as software. Sabre obviously did this for airline tickets. Email, I would even argue, did it for mail. True. Right?

00:05:20 Speaker_01
It's like you have files of mail, and now you just have a filing cabinet, if you will, on your computer. But the actions that were done on the software were really the same. So imagine an HR department that has 50 people working in HR.

00:05:33 Speaker_01
They might have one person that's in charge of the filing cabinets, and the HR person says, hey, get me David's file, because I want to talk to him about something, which is very scary. Don't worry. But get me David's file.

00:05:43 Speaker_01
So the filing cabinet person, the gopher, this is what the filing cabinet person would often be called, go for something. That's what it means. There's the gopher that obviously Bill Murray tries to kill in Caddyshack.

00:05:53 Speaker_01
But then there's the gopher of the person who goes for something, right? If you are a gopher at Creative Artists Agency, you are going and getting files. So that person went away. The filing cabinet went away.

00:06:04 Speaker_01
It was more efficient from a space perspective. But the 50 HR people are still 50 HR people today. So round one of software was really take the filing cabinet and put it not as physical files, but as a database with a front end.

00:06:17 Speaker_01
Round two, which started arguably at 1998, 1999, and this is what Salesforce did, the idea of a customer relationship management product

00:06:27 Speaker_01
has been around for a long time, or the Rolodex is an actual physical thing where you put every business card, organize them alphabetically, and you'd find the person that you want to contact that way.

00:06:35 Speaker_01
Salesforce put that not in software, but in the cloud. So what QuickBooks had done for a long time, or something called Great Plains Software had done for a long time, NetSuite did in the cloud. They put financial statements in the cloud.

00:06:50 Speaker_01
Zendesk put email support in the cloud. So it was still software, but instead of having a giant mainframe in your office somewhere, you now had it in the cloud.

00:06:59 Speaker_01
But again, going back to the HR example, so PeopleSoft first did this with on-premise mainframe computer, HR filing cabinet now is software. And then Workday came along, did the exact same thing, and now it's in the cloud. It's just much, much easier.

00:07:16 Speaker_01
Worry about your server like exploding in flames if your office building burns down or something. So it's more secure. So it was kind of software 1.0, then became software 2.0, which is in the cloud. And that played out from call it 1998 to maybe 2010.

00:07:31 Speaker_01
And then that grew a little bit with the insertion of financial services, because the way that I like to think about this is how many restaurants need and will pay tens of thousands of dollars a year for software?

00:07:42 Speaker_01
Like, Pan Am needed this in 1960, but does a restaurant with one location, are they going to spend $100,000 on a server and pay for software? No in 1960. No in 2000.

00:07:55 Speaker_01
But when the idea of bundling and payment processing became a thing in other financial services, now the market became big enough for restaurant software to exist. And this is where Toast came from. Toast is a $15 billion company that does this.

00:08:10 Speaker_01
Or Service Titan, again, what's the software market in 1965 for HVAC contractors? Zero. What's the cloud software market for HVAC contractors? Zero. But once you bundle on these other things, it becomes big enough.

00:08:21 Speaker_01
But the point that I'm getting to is that the same 50 HR people that worked in 1960 are the same 50 HR people in 2024. The same email support team in 2024 was the phone support bank in 1985, was the letter writing typewriter support team in 1965.

00:08:42 Speaker_01
What's exciting about AI is that it's taking this filing cabinet and now allowing actions on the filing cabinet. And that's what I think is really revolutionary, because you can actually ask the software filing cabinet application, like Workday,

00:08:57 Speaker_01
I want to add a dependent, and now Workday will do all the work involved with that. And Workday can charge a premium for that. So why is it input coffee and output code? The whole idea is that you now have software engineers that can build products

00:09:12 Speaker_01
on top of these cloud-based filing cabinets that now do the job that the end user of that software product did in 1960, 1970, 1990, 2000, 2010, 2023, 2024, but 2024, 2025 onward, now you have software agents that are effectively doing what for 65 years have been human work.

00:09:35 Speaker_05
Yeah, and that sounds really important, so I want to underscore that. You kind of went through the eras.

00:09:39 Speaker_05
You have the non-software era, then you have software, then you moved into cloud, then you have this financial services enabled cloud era, and now we're in this new era.

00:09:49 Speaker_05
Can you talk about how, as we chart through these different eras, the scale maybe in this new era is fundamentally different?

00:09:56 Speaker_01
Yeah, it's completely different because it's really comparing wages to software. And to take an example from a completely different field that also has no software market, there are about 4.7 million registered nurses in the US.

00:10:09 Speaker_01
Average wage for a nurse is a little over $120,000 a year. So it's a high paying profession. And that means that the annual nurse, not software market, but wage market, is over $600 billion a year. which is a lot.

00:10:23 Speaker_01
And the worldwide software market is under $600 billion, not just America. By the way, that nurse figure is just for the US. Of course there are nurses in the UK and France and Angola, like every country on earth, right?

00:10:34 Speaker_01
Probably Antarctica has nurses, I'm sure. Almost certainly. They're probably even higher paid down there. So the labor market is enormous, but what is the dedicated software market for nurses?

00:10:46 Speaker_01
Probably zero, because nobody took the time to develop software.

00:10:52 Speaker_05
The economics didn't make sense.

00:10:53 Speaker_01
Exactly. But if every hospital in the U.S. currently has a nursing shortage, or maybe you're in Minneapolis where there's a giant Somalian expat community and you need nurses that speak that language, how do you find them?

00:11:08 Speaker_01
And it takes three to four years to actually get trained as a nurse. Now you have a software product that can deliver not everything that a nurse does.

00:11:15 Speaker_01
Of course, it can't be a phlebotomist, it can't perform CPR, but it can call you the night before your colonoscopy and say, don't eat food. And it can say that in 45 languages, and it can actually have a conversation with you.

00:11:26 Speaker_01
So that's a labor example. Or going back to financial services land, NetSuite is used by, I think it's 70% of companies that go public are on NetSuite.

00:11:36 Speaker_05
It's some very, very large- I'm sure I've heard a podcast ad with that exact name.

00:11:39 Speaker_01
Yeah, maybe it's higher. Something on that order of magnitude. If I'm wrong by 10 points, it doesn't belittle the point. But the key thing is, you always have people that are paying you late.

00:11:48 Speaker_01
Like you'll look at your accounts receivable and you'll see, wow, a bunch of customers owe me millions of dollars. And that's what you'll see when you look at your financial statements.

00:11:55 Speaker_01
And then again, this is where humans take the operation on that information. I have teams that are in collections that will go call you and remind you to pay or I'm going to cut off the product.

00:12:05 Speaker_01
That's an operation that can now be done within NetSuite. They haven't done this yet, but I'm sure they will. And versus charging for the filing cabinet, now they can say, well, we know that you wanted to hire five collections people.

00:12:17 Speaker_01
We know that you pay those collections people $80,000 a year with benefits and everything else. We know that it takes a year to train them. Now our software product can do that not for $80,000 a year, but for $2,000 a year. And that's incredible.

00:12:30 Speaker_01
So the question is actually, how will the customer think about this? Because once they start paying NetSuite 10 times more than they paid NetSuite last year, they're like, wow, my software budget has ballooned. I got to cut my software spend.

00:12:43 Speaker_01
Or will they say, wow, I'm saving so much money. Because as opposed to these five job openings, I'm waiting to pay $400,000 a year for these five collections people.

00:12:52 Speaker_01
Now I can pay $10,000 a year to NetSuite, and I'm paying more for software, but less for labor. And that part, it's very, very new.

00:12:59 Speaker_01
But a lot of the hypergrowth that we're seeing in this category of company is because they are really moving into the labor market and less the software market. And that nurse example is a prime example of that.

00:13:09 Speaker_05
Definitely, it'll be very interesting to see the appetite if you think back to the app store early days when people were so reticent to pay 99 cents for an app for arbitrary reasons, but just because it was what they were used to paying.

00:13:23 Speaker_05
So as we round out these eras of clouds, you mentioned PeopleSoft or Quicken or Zendesk and all these companies that are capturing data. How are we uniquely set up now because of the previous eras?

00:13:35 Speaker_01
Yeah, so I would argue that if we just went straight to AI in 1960, this just wouldn't have worked because you still needed human input to go collect the customer information.

00:13:45 Speaker_01
We have everything built out where right now this isn't an API, but if I want to go answer a customer question, the question's already on the internet, it's already in a database. All of these things have set this up to be the ultimate platform.

00:14:00 Speaker_01
This is why we love investing in what I would call systems of record. And a system of record is just something that has every single piece of minutiae that runs a business. And it could be the strangest business that you can imagine.

00:14:12 Speaker_01
If I'm running a laundromat, there is actually dedicated laundromat management software.

00:14:17 Speaker_01
All of these systems of record that have popped up for all sorts of different businesses and all sorts of consumer use cases, like the fact that now the mainstream form of communication for many adults and children is texting and email.

00:14:30 Speaker_01
Like it's not voice, it's all in the cloud. Now you can perform these operations. So I would almost say there's been a 60 year period of digitization of physical things, putting them in the cloud. Why is the cloud part important?

00:14:45 Speaker_01
Well, again, if this were the mainframe era of the 1970s, how would you get the AI that's in some Google server somewhere? How would it actually have access to all of the information

00:14:56 Speaker_01
that's in some server basement in Indiana, like that's really hard to do. So all the data is here in the cloud, all the compute is in the cloud, and now you just mix them together.

00:15:06 Speaker_01
So the fact that systems of record for so many different types of businesses and so many different kinds of consumer use cases are now widespread,

00:15:14 Speaker_01
And hundreds of billions of dollars of company market cap has been created from these systems of record, either horizontal or vertical. And a vertical one would be something like a toast that vertically runs a restaurant.

00:15:25 Speaker_01
And a horizontal one would be like a Zendesk that just does customer support software in the cloud for every different type of company.

00:15:32 Speaker_04
Building on that, if you take the Toast example, start with the cloud wave, move to the financial services wave, it was initially a hypothesis that these vertical SaaS companies would make a lot more from financial services.

00:15:42 Speaker_04
Fast forward to today, 80% of Toast revenue is payments, insurance, all sorts of financial services versus software. And so if you're operating Toast, or one of my favorite examples is MindBody, which runs fitness studio software.

00:15:57 Speaker_04
It does scheduling for employees. It's a CRM. They also make a lot of money from financial services. But they still need a lot of people beyond the yoga instructor. You've got your financial back office.

00:16:07 Speaker_04
You've got people answering the phone to answer very basic questions. Like, all of that can start to be done with AI.

00:16:14 Speaker_04
And I think the most bullish version of that is, all right, you then don't need to hire people to do the tasks that are not human facing that AI can do better. Is that going to increase software revenue 2x?

00:16:25 Speaker_04
It could potentially increase it 10x, depending on how the customer views it and how much they're willing to let their software budget bleed into their labor budget.

00:16:32 Speaker_00
I think part of the challenge or the potential for disruption is that the pricing model may need to change pretty significantly, right? You talk a bit about this in your piece, Alex, where Zendesk today is charging on a per seat basis.

00:16:43 Speaker_00
But if you're actually eating away at some of the labor, you can charge for the output of work.

00:16:48 Speaker_00
And how does that create the potential both for increased ACV, but again, creates potential disruption for a lot of these larger incumbent players given their existing pricing structure?

00:16:57 Speaker_05
Can we actually talk about that example from Zendesk? What is the difference between the software component and the human component?

00:17:03 Speaker_01
Well, it shows how stark the difference is. So most companies, like Salesforce charges per seat. And so Zendesk, at the time that I wrote my piece, they might have changed their pricing a little bit, but it was $115 per seat per month.

00:17:17 Speaker_01
So imagine that you have 1,000 people that work in an email-based support center, and they use Zendesk. And Zendesk then profits as you hire more people, because 1,000 is better than 100. So $115,000 a month is about $1.4 million a year.

00:17:34 Speaker_01
in spend on software for Zendesk. And Zendesk, I think, has about $2 billion in revenue, something in that order of magnitude of annual recurring revenue from all of these seats that are paying every single month.

00:17:45 Speaker_01
And of course, they want their customers to grow seats. Now, if you assume that each seat, how much is that person paid? How much does their healthcare cost? How much does their stipend for commuting and yoga benefit?

00:18:00 Speaker_01
Maybe it's $50,000 a year per person. So what's 50,000 times 1,000? Again, 1,000 seats, that's $50 million. So you're spending $50 million on people. And by the way, it's very hard to hire and train these people.

00:18:13 Speaker_01
So I think one misnomer is like, oh, AI is terrible. It's going to end all employment, and it's going to be chaos. I always like to point out that when the United States was founded, 97% of people in the US were farmers.

00:18:26 Speaker_01
And most of them were put out of jobs by things like the tractor, but that's actually fine because they moved on to other kind of productive parts of labor. And by the way, the average life expectancy was a little over 30 back then.

00:18:36 Speaker_01
So I feel like things have gotten better. I like penicillin and all the other benefits that we've gotten, and fertilizer and all these things that have made our lives better.

00:18:44 Speaker_01
So you have $50 million a year for people, and then you have $1.4 million a year for software. Which one's bigger?

00:18:50 Speaker_01
And this is the concern is that on the intermediate basis, and Zendesk is actually very lucky because they were a public company, and they got taken private. So two big private equity firms bought it.

00:19:01 Speaker_01
And they're actually working on this right now because they're like, uh-oh, if we make AI really good, then the customer that has 1,000 seats might cut down to 10 seats. Because there are two forms of AI tools.

00:19:16 Speaker_01
There's more than two, but the common example that we talk about is there's autopilot and then there's copilot. So copilot is a productivity enhancer. So I'm trying to figure out how do I answer Angela's query? I just got hired yesterday.

00:19:30 Speaker_01
I don't even know where the bathroom is. What do I do? And then it's like, hey, we think this is the right answer. So it makes Angela so much more productive in her job, and that's great.

00:19:39 Speaker_01
Autopilot is like, Angela quit yesterday, and we need somebody else to go answer emails because it's Black Friday. What do we do? What do we do? OK, now we just throw the tool at the customer directly and have them answer the questions.

00:19:53 Speaker_01
And that's the big danger. Copilot is actually a danger for revenue as well, because why do I have 1,000 support reps? Because they can only answer 10 questions a day. And I get 10,000 queries a day, so it's just basic math.

00:20:05 Speaker_01
Now with Copilot, each one of my reps can answer 100 questions a day. I only need 100 reps. So now Zendesk lost 90% of its revenue. And then by the way, if autopilot becomes a thing, and it actually works very, very well, then I need nobody.

00:20:19 Speaker_01
And therefore, I sell no seats if I'm Zendesk. So it's both defense and offense for these companies to figure out what the hell to do.

00:20:27 Speaker_01
Because if they play offense, they could maybe 10x the revenue because of the example, like it's $50 million for people or $1.4 million. I'd rather get $50 million from Zendesk than $1.4 million.

00:20:37 Speaker_01
But they're probably not going to be able to take all $50 million, right? Because ultimately, the cost of delivering these services is very low. The moat is not that high.

00:20:46 Speaker_01
It's higher for companies that have the system of record, I would argue, because all of your past correspondence with all of your customers is in a Zendesk.

00:20:54 Speaker_01
Where if I'm Salesforce, like every single communication that I've ever had with any of my customers, my pipeline, everything, it's in Salesforce. It's hard to yank that stuff out because everybody's using it.

00:21:03 Speaker_01
And it's not like this binary thing where tomorrow we're all autopilot. We're going to see a lot of these copilot tools. They will run on the systems of record. But even as they're running out of systems to record, now I need less seats.

00:21:15 Speaker_01
That's why Salesforce, it's a $200 billion plus public company. If they don't do this right, they could lose all of their revenue, or most of it. If they do it really well, they could 10x their revenue. And it's like, where is it going to go?

00:21:26 Speaker_00
As early stage VCs, we're very excited about this. Yes, yes, I know. It's all up to us.

00:21:29 Speaker_05
This is a question, right?

00:21:30 Speaker_04
We worry more about finding the next entrepreneurs versus the aid of the incumbents.

00:21:33 Speaker_00
It's an interesting moment in time for enterprising young founders to reinvent the model and charge radically different. evolve. They're not going to change their perceived pricing and they risk disruption.

00:21:44 Speaker_05
And is that the wedge, right? If you're a startup and you're trying to figure out how do I enter the market, especially when you have companies like Salesforce, which have the system of record, they have all the data, they have all the customers.

00:21:55 Speaker_05
Is that the wedge where you say, I'm going to undercut and I'm going to charge a tenth of the price, even though I still have great margins because I'm entering the labor part of the equation?

00:22:04 Speaker_00
I think so, yeah. I wrote a piece recently that I called the messy inbox problem, which is my way of describing a wedge strategy that we're seeing across lots of different industries.

00:22:12 Speaker_00
And the idea is basically that there's a class of founders that are building software products to solve a lot of the what was historically judgment intensive work in lots of different industries. There is some sort of human administrator's job.

00:22:25 Speaker_00
It is to basically extract information from a wave of unstructured information, whether it's emails, faxes, transcribing phone calls, and then put that information into one of these downstream systems or record.

00:22:37 Speaker_00
It could be an EMR, it could be an ERP, it could be a CRM system. And historically, that work lived upstream of any of that software, right? Because it was the human's job. And software couldn't do that.

00:22:48 Speaker_00
Now we're seeing companies sort of wedge in and replace, again, that messy inbox problem with software and slowly begin to eat away at all the kind of downstream workflows.

00:22:57 Speaker_00
And over time, I think that the thesis is that while that initial wedge is highly differentiated against the human, it really is the opportunity to eat away at everything else and become the kind of new AI native system of record.

00:23:10 Speaker_00
We have a company as an example called Tenor that is doing this in a healthcare context. So the problem that they're solving specifically is around patient referrals.

00:23:17 Speaker_00
So you go to your general practitioner, they're referring you to a specialist, could be a dermatologist or an imaging center.

00:23:25 Speaker_00
often faxing your medical records, and it is somebody's job to go physically to the fax machine, re-enter that information into the EMR system.

00:23:34 Speaker_00
Tenor has trained a model against, I think, 4 million healthcare-specific documents, and now can basically extract all that information about the patient programmatically, and have effectively begun to solve this patient intake problem.

00:23:47 Speaker_00
And they're able to reduce now about 90% of the admin costs. of that patient intake before the patient's actually seeing the clinician. So they, again, wedged in with the messy inbox problem.

00:23:57 Speaker_00
And over time, they're now eating away at things like scheduling and eligibility and benefits. And over time, we'll see if they become the kind of core AI native system record.

00:24:05 Speaker_05
And something you're pointing to there is also the defensibility of it all, right? So you can get the wedge maybe through pricing, but how do you actually protect your customers?

00:24:13 Speaker_05
I think one place people jump to is, okay, you need your own models, or you need some sort of proprietary data. Is that the defensibility of this future, or how do you think about that?

00:24:23 Speaker_00
It's sort of this distinction between differentiation and defensibility. I think AI is an incredible catalyst for differentiation, right? Solving the messy inbox problem with software is a thousand times better than the human doing.

00:24:34 Speaker_00
It's not even on the same kind of playing field. Super differentiated way to wedge in, again, kind of own the downstream workflow. Is that wedge product alone defensible? I would argue no, right?

00:24:45 Speaker_00
Today, it feels like magic to the providers that they're working with. But I think that capability is going to become commoditized over time. They may have an advantage because they've trained a model for now, but I think that is ephemeral.

00:24:56 Speaker_00
I think the defensibility comes from, again, owning all of the downstream workflows, deeply integrating themselves into every other system that they have, effectively owning that sort of core kind of end-to-end workflow.

00:25:09 Speaker_00
And I guess the hot take would be that moats still matter. And a lot of the moats in software today are the same that they've always been.

00:25:16 Speaker_00
So becoming a system of record, having a network effect, becoming a platform, having virality baked into your product, deeply embedding yourself into the existing system so it's hard to rip out, like these were all of the heuristics that we would always have looked for in software.

00:25:31 Speaker_00
And they're still true today.

00:25:32 Speaker_01
I agree with all of that. The other way to think about this is, why did software start with airlines? Well, a lot of people traveled. Airplane tickets were very expensive, and this was a pittance for them.

00:25:45 Speaker_01
And it made so much sense versus having throngs of filing cabinets and gophers, and it makes sense to go pay. hundreds of thousands of dollars in 1960s money to go buy some giant IBM mainframe.

00:25:57 Speaker_01
And this is why I brought up that financial services example. Software for restaurants did not make sense. There just wasn't a problem to be solved and the market wasn't big enough.

00:26:07 Speaker_01
And you made the market big enough once to Angela's point, you threw in like payment processing and insurance, all these other things that they were paying for anyway.

00:26:14 Speaker_01
So you can either try to come up with a wedge, as David mentioned, and then figure out how you expand your wedge and make it defensible and become the system of record.

00:26:22 Speaker_01
The other thing that you do is you just find things like restaurants in 1980 that have no software and needed no software, wouldn't pay for software, but their labor budgets are enormous.

00:26:32 Speaker_01
And the example that I would give here, what is the incumbent software product for compliance officers at banks and financial institutions? Excel? Word, Microsoft Edge browser looking for bad things. What does a compliance officer do?

00:26:47 Speaker_01
I mean, it's now in the news a lot because of this debanking thing. And I found this on the Bureau of Labor Statistics. The fourth fastest growing job in America is compliance officer. There isn't an incumbent software product that they use.

00:26:59 Speaker_01
Every bank and financial services company on earth, they're all hiring for compliance officers. It takes a long time to train them. And what if account openings goes down? I don't need as many.

00:27:10 Speaker_01
It takes in some cases like a month to open a business bank account at a bank because the compliance officers are backlogged. What if you deliver that via software? And there is no software.

00:27:20 Speaker_01
There is no incumbent that can now add this AI module in the same way that NetSuite was an incumbent for accountants and financial officers at companies where they can add a module for AI work of collecting money.

00:27:32 Speaker_01
So you find these other areas where there really isn't an incumbent or the incumbent is like Microsoft Excel. And sometimes these are just so bizarre.

00:27:41 Speaker_01
Like I wouldn't think about compliance officer until I saw this random report from the Bureau of Labor Statistics that showed manicurist is number one. And that's a hard one to have an idea. And then number four is compliance officer.

00:27:53 Speaker_01
And then you talk to banks, like, what software do you use? And it's Excel. Again, giant labor budget, not enough people that are using software. You can not have to worry about the incumbent.

00:28:03 Speaker_01
And it still is a wedge, but you could probably turn it into a system of record. And why we're no venture-backed companies or even non-venture-backed companies built in this space, it's just like, well, I can't charge that much money.

00:28:13 Speaker_01
It's just not a big market in the same way that there was no restaurant software market in 1980. It's the exact same reason, but there's enough budget there to go fill this very, very pressing need.

00:28:23 Speaker_04
I think one of the most interesting parts of our job is every time there's a new technology shift, we have to challenge every investment thesis that we thought was not going to work. And so many of them now are going to work.

00:28:33 Speaker_04
If we come back to financial services, there's a lot of pretty terrible systems of record where smart people have tried to get them ripped and replaced, and it was just not going to happen.

00:28:43 Speaker_04
And my new conclusion with AI is not that they would never do it. It's that the replacements were 2x better. They weren't 10x better.

00:28:50 Speaker_04
And so if we come back to compliance, lots in Excel, but for instance, you've probably read the $4 billion fine by TD in transaction monitoring, right? And so they said an old transaction monitoring system maximizes one of them.

00:29:01 Speaker_04
They should probably get one that throws up fewer false alerts, but they were trying to clear a backlog of several tens of thousands alerts. They can't hire enough compliance people.

00:29:12 Speaker_04
So now an interesting wedge in is we'll provide you with all of these agents. And oh, by the way, we also have a much better transaction monitoring system that is actually going to fix the problem.

00:29:21 Speaker_04
So this labor plus software bundle also helps the sales process and then helps the defensibility because you're really solving the major problem, which is better software and also I can't hire the people.

00:29:32 Speaker_05
Yeah, and I think that's such an important point because you all have pointed out different areas where, quite frankly, the labor is not there.

00:29:39 Speaker_05
And so as we're talking about software disrupting labor, the natural question is, OK, so what happens to all these jobs? But maybe we can talk about both that and the flip side of what new jobs are created.

00:29:49 Speaker_05
Because the previous arcs, we saw product managers, UX designers, social media managers. Those were all remnants of the previous era. How do we think this will shape up?

00:30:01 Speaker_01
It's always hard to prophecy these things because in 1789 or something, it'd be hard to say, what will these farmers be doing post-tractor?

00:30:08 Speaker_05
Sitting in a room talking.

00:30:09 Speaker_01
Exactly. Talking with these electronic microphones and this like amazing fire. Look at the fire in the sky here that we have in the basement.

00:30:16 Speaker_05
Our ancestors would be proud.

00:30:17 Speaker_01
Exactly. You know, what was a nurse, right? It's like medicine was bloodletting leeches and prayers. So it's obviously changed a lot.

00:30:24 Speaker_01
So on the one thing that I think AI cannot do, and in fact, if you have an AI sales rep, like Salesforce, why do I need seats for salespeople if AI is doing selling? But AI cannot build a relationship with somebody over golf.

00:30:37 Speaker_01
So I think the in-person things that only humans can do, that skill set might go up in value tremendously. At the far extreme, I talked to somebody who believes that in the distant future, there will only be two jobs.

00:30:49 Speaker_01
You either tell a computer what to do, or you were told by a computer what to do.

00:30:53 Speaker_01
And there's a whole set of things where people could be a lot more productive in whatever job they're doing when you have this little coach by your side saying, do this, do that.

00:31:02 Speaker_01
But I think the human connection thing is almost the most important because if I think about every other era of a new communications tool, imagine having the first telephone. Alexander Graham Bell invents the telephone. Nobody has a telephone.

00:31:14 Speaker_01
Like, how do you scale this network? You get one. And the only person that calls you is your mother saying, why don't you call me more often or something?

00:31:20 Speaker_01
So then the first telemarketer shows up and then starts taking advantage of the fact that you have a phone and they can, rather than go horse and buggy to your house, they can try selling you something over this like old fashioned telephone.

00:31:33 Speaker_01
That was a very advantageous place for that first telemarketer to be. Or the history of Sears Roebuck is really fascinating because even though they had the Sears Tower, they had these giant stores, it really was the first giant mail order catalog.

00:31:47 Speaker_01
So they were the ones that kind of figured out how to use the US Postal Service. Faxes came out, people started sending unsolicited faxes.

00:31:54 Speaker_01
So the reason why I bring all this up is you can imagine a world where like AI is selling you everything, pushing you everything, and then right now it's a novelty and it works really well.

00:32:02 Speaker_01
But once it becomes so mainstream that everybody's doing it, it's like this Yogi Berra expression, like it's so crowded nobody goes here anymore. You can imagine like the need for actual human connectivity to go up dramatically.

00:32:13 Speaker_01
It's followed this pattern of once it gets so crowded, somebody that's doing something different, in this case the old-fashioned way, it might be more valuable.

00:32:20 Speaker_04
One of the ways we think about it is all of us have some percentage of our job that's rote tasks that could be automated with AI. And I think we strongly believe that at least every white-collar job is going to have a co-pilot.

00:32:32 Speaker_04
Some might be fully agentic, back to our L1 compliance reviewers.

00:32:36 Speaker_04
And so if you imagine all of us not doing any menial tasks or anyone focusing on the human connection, the most creative parts, and having all of our days to spend on that, like what might be enabled?

00:32:47 Speaker_04
And that's a pretty exciting way to think about it.

00:32:49 Speaker_05
Definitely. And as we think about the companies that can be created in this wave, I'm so curious because it does feel fundamentally different as you three are assessing companies. Are there a new layer of metrics that you pay attention to?

00:33:01 Speaker_05
Again, using a previous wave, maybe we got social media and all of a sudden we're thinking of daily active users. That was a key metric that people started to pay attention to. Are there new metrics? Are they the same metrics that matter?

00:33:12 Speaker_05
Is it too early to tell?

00:33:14 Speaker_01
I think it's actually the exact same metrics. It's not like, oh, it's AI, so therefore future profits don't matter. It's the present value of future profits. And that really comes down to how many customers do you have? Do you retain those customers?

00:33:25 Speaker_01
And then how much gross profit do you make per customer? And then how much overhead do you have? And I don't think any of that changes. So I mean, the reason why social networks were interesting is we knew that the customers retained.

00:33:38 Speaker_01
But will people pay for it? Will it make money? There was this open question. And therefore there was this, if you will, alpha of, oh wow, we call this the smile curve, which is very rare.

00:33:48 Speaker_01
Obviously 100% of people use the product on day zero, because day zero is when they installed it. But then people stop using it on day one, day two, day three. And normally most products, they just have exponential decay.

00:33:58 Speaker_01
So by day 200, of the 100 people that downloaded it on day zero, zero people use it at the end. What's interesting is things like Uber or Facebook, where, again, 100% of people use it on day zero.

00:34:09 Speaker_01
Then it drops off on day one, day two, and then it picks back up. And then it plateaus at maybe 50%, 70%, 90% of the original starting bunch. And that's so rare. But then the question was, will Facebook ever make money?

00:34:22 Speaker_01
Oh, they won't make money because it's free. But they figured out that advertising was very valuable. I think the vast majority of things that we're seeing right now just monetize via subscription. So it's actually very clear how they make money.

00:34:33 Speaker_01
And the DAU thing is actually just as useful today as it was before, but the money part is almost automatic. Like the thing that was unique about the internet era was it's like, oh, get big and then monetize later.

00:34:45 Speaker_01
And we're not seeing as many of those, but I don't think any of the fundamental isms of evaluating a business have really changed. The only thing that's more dangerous is since AI can now write software,

00:34:56 Speaker_01
It's just so much easier to spin these things up, whereas to build something and scale it out, the reason why Friendster failed, like Friendster should have been the social networking winner, but their servers couldn't stay up.

00:35:09 Speaker_01
And MySpace should have been the winner, but they couldn't hire good engineers. There are all these different reasons that are not relevant today because the technology stack is so different. But again, it's present value, future profits.

00:35:20 Speaker_01
That's unchanged.

00:35:22 Speaker_04
I think one thing, and this is not a business metric, but that's changed is just potential market size. And we talked about how it scales in the large vertical SaaS markets. So there's something in the U.S.

00:35:32 Speaker_04
called the North American Industry Classification System, the NAICS codes. And there's 600 of them, and they'll classify industries, how many companies are in there, what's their labor budget.

00:35:40 Speaker_04
And there's a whole host of industries, whereas if you looked at them before, you say, well, there's 1,000 potential buyers. Maybe they'll pay $1,000 a month for my software service. That's a $120 million market.

00:35:52 Speaker_04
That's really not that interesting if I'm going to build a venture-backed business. Now, if you think you can layer an AI, replace some of the labor budgets, those markets get dramatically bigger.

00:36:03 Speaker_04
And so I think the different pockets of where software can be built, these niche markets that weren't that interesting, are now potentially very interesting.

00:36:10 Speaker_00
I think the other dimension that we're seeing pitches for is, are you selling software into the incumbent industry, or are you building the full-stack version?

00:36:17 Speaker_00
Alex wrote a bit about this in Herbarians at the Gate, which is sort of like the evolution of private equity kind of in an AI context.

00:36:24 Speaker_00
And so you think about an area like professional services or in legal, for example, the challenge that a lot of these law firms have is that they're charging on a per hour basis, right?

00:36:31 Speaker_00
So if AI can do what used to take three hours and three seconds, where does the revenue go? And so we're seeing some people pitch

00:36:38 Speaker_00
the full stack kind of AI native law firm, which might have a totally different cost structure to Kravath or one of these big firms. Or there are other areas within professional services that are much more aligned to benefit from that efficiency.

00:36:50 Speaker_00
So for example, we have a company that is solving a lot of the workflow challenges in plaintiff law.

00:36:55 Speaker_00
They operate in both employment and personal injury, where in that model, unlike on a per hour basis, they're charging on a contingency model, meaning they don't get paid unless there's an outcome or a settlement in the case.

00:37:07 Speaker_00
And in personal injury, as an example, for every 100 leads that these lawyers get, they take one case.

00:37:13 Speaker_00
And there's a ton of, again, that messy inbox problem of sifting through medical records or employment documents and having to essentially quantify the value of each case that they'll take on, because any case that they take is an investment of their labor.

00:37:25 Speaker_00
And so what this company is doing is essentially programmatically helping solve that sort of intake challenge, that messy inbox problem, to help automatically qualify the value of those cases.

00:37:35 Speaker_00
And then works essentially as a copilot for the lawyer to draft a medical chronology, to create a demand letter, to file a complaint, and basically walks through the entire sort of pre-litigation and litigation process.

00:37:48 Speaker_00
That allows that lawyer to take on 3x or 4x the number of cases. But again, the value that software is delivering to the practice is reducing labor costs. So one way to do it is you just have fewer lawyers and the same amount of revenue.

00:37:58 Speaker_00
Or in this case, I think what's going to happen is it's going to significantly grow these practices. And in this case, they're actually passing the cost of that software to the end client.

00:38:08 Speaker_00
in the form of a technology expense, which they often had done historically. And so the value that the software is delivering to each of these firms is super aligned to the impact that it's having on the business.

00:38:18 Speaker_00
And so I think the more clients you can take on, the more people that can pay for the software. And so on a per firm basis, there's a significant kind of revenue expansion opportunity.

00:38:26 Speaker_00
I think that's an interesting tension that you'll see across industries where does the AI help by reducing cost? Is it better to build a full stack version or sell the software in?

00:38:35 Speaker_00
And I think there'll be successes in both dimensions, but it's something we are seeing more of.

00:38:40 Speaker_05
Yeah, and one follow-up there, as you're talking about the cost being passed along to the end user or buyer, is that just net deflationary as this permeates across the system?

00:38:51 Speaker_05
I know it would take time, but eventually, if you see more competition and more people creating these AI-based labor products, And then people are competing on price and all of a sudden to take on a new case, it's no longer $5,000, it's $500.

00:39:03 Speaker_05
Is that something that you guys are thinking about? Or you know how you said that in the previous era, the open question was, can we make money? Can we monetize?

00:39:13 Speaker_05
Is that an open question that this, over time, just becomes deflationary and firms can't charge as much?

00:39:20 Speaker_00
I don't know, this legal example might be unique in the sense that the clients are also waiting for a settlement themselves. And so the cost of this offer is essentially coming out of whatever the lawyer is able to help win for the client.

00:39:31 Speaker_00
And so they're not really feeling the cost of the technology as much as maybe other industries.

00:39:36 Speaker_01
I think technology, if it's done right, is always deflationary because you get productivity gains.

00:39:40 Speaker_01
So I think for sure, I think the defensibility point is the most relevant one that we struggle with a lot, which is, wow, it's so easy to build one of these things. Because the number one use case, it's almost like a recursive thing.

00:39:54 Speaker_01
It's like, which profession is using AI tools the most? It's probably the tech people that actually build tools. And that's where things like Cursor have gotten so popular.

00:40:05 Speaker_01
Companies like Stack Overflow are suffering so much because I don't need to go to Stack Overflow anymore because I just get the answer from Claude or I get the code sample from Cursor. It's just so much easier to do this stuff.

00:40:18 Speaker_01
And so it is fundamentally deflationary. But what if there are 50 companies that end up doing the exact same thing? That's the hard part.

00:40:24 Speaker_01
But I think I can't see a scenario where prices are more expensive than humans or where prices don't just keep going down.

00:40:31 Speaker_05
And significantly.

00:40:32 Speaker_01
Significantly. And that's the history of technology in a nutshell. I mean, again, it's like the 100 megabyte hard drive in 1960 that weighed like tons, literally tons. It was probably a million dollars or something crazy. And now it's just comical.

00:40:45 Speaker_01
I bought some for Cyber Monday. I bought all these little micro SSDs for a terabyte was like $10. It's just incredible. So I think that is an inexorable process for technology cost in general.

00:40:57 Speaker_01
And then you also get new use cases where I've written about this separately, which is the everything to the right of the supply-demand curve. This is a really interesting use case where it's like there just wasn't demand where there was supply.

00:41:10 Speaker_01
There's a lot of supply to do something for $2,000 an hour if I want to file a trademark with the leading trademark attorney. So the trademark market or the patent market, like that might be very small because it costs too much.

00:41:24 Speaker_01
But now if it only costs $5, wow, maybe everybody does it.

00:41:27 Speaker_01
Or like translation is one that I find fascinating where it just doesn't make sense if you're a small company to go translate your introductory video into like 45,000 different languages that have ever existed.

00:41:39 Speaker_01
Like, why would I translate it into ancient Greek? But you know what? Why not? It's free, right? So you have all these other things that just expand the market because the cost has dropped so precipitously.

00:41:51 Speaker_05
Closing things out, where do you guys want to see more builders applying themselves? You're obviously seeing a lot of companies, a lot of people excited. Also, the incumbents are clearly excited about getting in on this wave.

00:42:01 Speaker_05
Is there an area that you'd like to see more attention being put toward?

00:42:05 Speaker_01
My view is obscure is good. We love it when somebody walks in, has had a decade or more of obscurity. They serve some weird job, or they were at something that nobody's ever heard of, the farming industry, the mining industry, the whatever industry.

00:42:20 Speaker_01
And then they actually have an insight that somebody else doesn't, and they actually know the potential of AI. Because the other thing is that it's very important to know that the technology is not ready for autopilot for a lot of these things.

00:42:32 Speaker_01
It's just like the use cases are too complicated. The integrating the different pipes is too complicated. Overshooting early, like there are going to be a lot of failures because there inevitably are in every technology revolution.

00:42:42 Speaker_01
Not because the idea is bad, but it's not good enough to be a hundred times better. I think it's like obscure and then finding the ones where like, at least for right now, actually the technology is good enough for the obscure use case at hand.

00:42:55 Speaker_04
Yeah, I'd say also across, there's many industries of this, but financial services and insurance have a host of old systems, like 30 plus year old systems of record that now can be made 10x better, incorporating labor, redoing them in a workflow.

00:43:09 Speaker_04
And so deep knowledge of those areas, like we have transaction monitoring in Sardine, we've got a mortgage loan origination system in Vesta, we've got servicing, we've got a couple in insurance.

00:43:18 Speaker_04
And so entrepreneurs that really understand those space and can bring AI thinking there, I think is a big opportunity.

00:43:25 Speaker_00
I think we'll continue to see lots of entrepreneurs wedge in with the messy inbox problem across lots of niche vertical industries.

00:43:30 Speaker_00
We're still on the lookout for horizontal software, AI native versions selling into sales teams, marketing, product management, analytics, CFOs. In those categories, you often do have a large incumbent software competitor.

00:43:43 Speaker_00
And so that's sort of the tension. You have to understand the market structure and how likely it is for that incumbent to change their pricing model and build more AI-native features.

00:43:51 Speaker_00
But I think there will be generationally defining companies built in an AI-native way in horizontal software as well.

00:43:57 Speaker_03
Very exciting. Thanks, guys.

00:43:59 Speaker_00
Thank you.

00:44:00 Speaker_03
Thanks, moderator. You would have been a good farmer, too, I'm sure.

00:44:05 Speaker_05
All right, that is all for today. If you did make it this far, first of all, thank you.

00:44:10 Speaker_05
We put a lot of thought into each of these episodes, whether it's guests, the calendar Tetris, the cycles with our amazing editor, Tommy, until the music is just right.

00:44:18 Speaker_05
So if you like what we've put together, consider dropping us a line at ratethispodcast.com slash a16z. And let us know what your favorite episode is. It'll make my day, and I'm sure Tommy's too. We'll catch you on the flip side.