Vertical SaaS is Dead. Long Live Vertical SaaS.
AI, business models, and paradigm shifts
Vertical SaaS, and really all of software, has long been a game of applying efficiency gains to industry-wide problems in the hopes of capturing a cut of the gains via a recurring SaaS contract.
One could easily predict this to continue. And in fact, early AI products have seemingly pointed in this direction. The business model for vertical SaaS after all has been predicated upon diffusing and codifying industry best practices in software. If an industry’s best practice can be made more efficient via AI, perhaps we can simply apply AI to existing processes, make them more efficient, and continue on with the SaaS licensing model.
And I’m convinced this will continue, in no small part, because there are several forces working in tandem to guarantee this: First, many industries have no real desire to disrupt their current business model. Second, in the event business models do get disrupted, there will be SaaS vendors waiting to enable this. Third, the marginal cost of software may approach zero, and yet it doesn’t matter in vertical software. Many industries have no effective R&D budget to take advantage of it. They will still pay a vendor to manage their technology stack. Think I’m wrong? Explain to me continued prevalence of the MSP industry then! There will always be incredible pure play software opportunities. There may just be far fewer venture-backable ones.
Because there does exist a future where this is flawed thinking. After all, there’s a foundational layer necessary for the vertical software play to work: the market must be sufficiently fragmented.
And this strikes me as potentially the largest problem for positing the continuation of the vertical SaaS business model in the new world. It’s my belief, that there are many industries where the right play is to simply become the AI-powered consolidator.
What’s been left on the table for the next wave of vertical companies is to make their business, the business of business model disruption.
Amazon
What was the better path to value creation for the entire stack of consumers, employees, shareholders, and more in 1994? Amazon or vertical SaaS for bookstores?
There’s one right answer here. The Internet was a platform change. And certain businesses got blown out in favor of new consolidators.
The Internet birthed a new platform with new reach to the consumer and new ways of supporting this unprecedented consumer reach.
Obviously this only really mattered for businesses that could support an exponentially increasing amount of customers without scaling opex proportionally.
For every other industry, the internet impacted their distribution strategy - every SMB has a social media strategy now - but it didn’t impact their opex or COGS. Your accounting firm has needed to staff proportionate to the amount of businesses they are working with. Same with law, same with most service sectors.
The key platform change inherent in AI is not around distribution, but around the scaling costs of opex. If AI succeeds in increasing throughput in a business, this in turn invites the possibilities of the Internet and consolidation into new industries.1
AI invokes an arms race. Businesses must grow or die in the age of the internet. Offline industries have mainly grown inorganically or not at all.2 The former peace deal in fragmented industries was brokered around the margins required to service customers. It didn’t leave a ton for hefty growth targets. The prospect of AI has effectively nuked the peace treaty. Margins in knowledge work domains are expanding. The only question that remains for industry participants is: so then what are you going to do about it? If your costs plummet to serve customers, sure, you can take a nice profit distribution, funnel none of these increased profits back into growth, and ride out into the sunset. However, my hunch is this would only last about 3 years prior to a new competitor waiting to take market share away. Your margin is my opportunity. And if the costs to scale are way less, somebody will do so. The arms race is on.
Okay, but why not Pure SaaS?
“To get things started, the company that introduces an innovation usually builds all the complementary pieces required for that innovatoin to function. The bundled technology stack - trains, tracks, train stations, rail yards, signaling equipment, and so on - is an essentially unavoidable hurdle for first movers.”
Mapping the Space Economy, Chad Anderson
The extent to which you agree with this thesis is going to be dictated by two elements:
Is AI a product, a paradigm, or something else?
Will traditional private equity win consolidation and innovation battles?
We are Pre-Paradigmatic
To answer question one, I think the easy answer is that we are effectively still pre-paradigmatic. This is a result of a couple things:
All productivity increases are flowing primarily to the employee and not the company right now.
Working with LLMs in a deep way requires sociotechnical reconfiguration of organizations that incumbents are generally fearful of undertaking.
Many complementary technologies in industries are still incredibly dysfunctional in verticals, lessening the impact of AI altogether.
In other words, this isn’t a pure tech issue. It’s a far deeper paradigmatic shift that needs to be undergone for AI as a product to be talked about meaningfully. In the short to medium term, instead all the value flows to organizations that primarily use finetuned, industry specific LLMs as an internal tool. Until that happens, good luck on the pure SaaS play for AI.
This is worth spending time on more deeply, but will have to wait for a better time. The paradigm hasn’t shifted and foreseeably won’t shift to make LLMs productizable until the first consolidators start to meaningfully enter industries with this thesis.
The latter point is far more interesting as it pertains to investors. There is deep skepticism that existing PE companies won’t effectively win the consolidator play.
Does PE eat your lunch?
The objection goes like this:
PE as a whole is really good at acquiring businesses - like really good.
They’re even quite good at value creation.
The PE shops will simply utilize AI inside their businesses and consolidate even faster.
My response goes something like this.3
PE upper-mid market and above is in a great spot as they will be acquirers of the assets tech-enabled consolidators pull off. Strategics as well. The lower middle market is in a tough spot.
Effectively, the people whom would be first movers on any interesting acquisition or franchise play will not be. The initial capital expenditure doesn’t de facto fit into their theme or their skillset: AI R&D teams and paradigmatic restructuring around it.
They are simply on equal ground in the 0 to 1 phase in any industry where AI gives a compelling tailwind, but requires a certain capex spend initially. Nor is this necessarily a mandate that their LPs give them.
Second, they can’t simply purchase the tech off the shelf to implement and realize margin improvement. To reiterate my view, AI as a product can’t meaningfully exist until the first movers move. Every one is on equal footing. The PE acquirers of certain assets are no better off than the startup.
As such, we are simply looking at stasis in PE to move on these strategies until the paradigm is meaningfully shifted.
Well okay, even if you buy this, what do the returns look like? Jury’s out! Perhaps it’s worth studying the IRR of the best franchise concept groups which are quite impressive. Or the Constellation strategy as it applies to extremely underpriced assets in the AI era. Or Wayne Huizenga. Many of these had a paradigmatic advantage that then exercised resulting in massive amounts of value creation and multiple expansion that defies what PE is thought of as doing.
I strongly believe this is what will happen. And there are a new crop of Wayne Huizengas and Mark Leonards that will create billions of shareholder value at high growth rates.4
The AI paradigm also requires a new sort of investor paradigm, but I think this is a theme that LPs that invest across asset classes will find very exciting.
If you know/are one of these investors, please hit me up! There are very few of them today.
What comes next?
So what does this mean for vertical SaaS? Well, I think we must start with the assumption that the AI arms race will kick off in every industry leading to consolidation. And you must then figure out how you are going to operate as a result.
Yoni Rechtman and the rest of Slow Ventures as well as Forerunner have been articulating some of the new business models that seem to await the new world. These namely look like a) digitally native franchises or b) vertical SaaS buyouts.
I’ve written about these models, so I won’t be spending much time revisiting the models themselves.
Instead, I think it’s more interesting to think through when to apply either model. When is it best to buy existing industry distribution vs. create new distribution altogether?
The key to answering this is a convictional one: Can you catalyze new business formation inside of an industry that wins distribution? Or is distribution captured in the form of long term contracts or relationships? This is the key difference vs. the Amazon strategy. The internet was so clearly a distribution innovation that it only made sense to build wholly anew. Other industries require more thoughtfulness.
Historically, vertical SaaS investors and operators have looked at industries with low new business formation and assumed that new business formation is really difficult to create. I think that’s wrong. New business formation is a response to opportunity latent in an industry for margin capture. Margin expansion, outdated business models, and a wave of young people whom don’t see the possibility of career advancement in the existing model means new business formation is now attractive in most industries. It’s simply waiting on someone to catalyze it.
New business formation and total factor productivity go hand in hand. So long as you can actually expand margins and productivity inside of the industry activity, all industries are now worthwhile for a model that seeks to catalyze new business formation.
As far as figuring out what mechanisms catalyze the distribution? There are secrets everywhere.
But there are certain industries which already lean towards consolidation. As an example, where the costs to service and staff a contract are large enough, a certain amount of consolidation will already be present. What these existing consolidators largely haven’t done is focus on the consumer experience as a way to reduce costs to service. If you can improve that, you have a distinct margin advantage that can be further increased via AI in the back office.
I think the brilliance of Metropolis, the pioneer here, was in focusing their solution on the consumer experience which then necessarily implicated a change in the business operations.
Not a single one of the parking giants could figure out how to meaningfully drive better parking experiences. And it wasn’t for lack of trying, they all acquired technology solutions to try.
Perhaps the calculus here for vertical SaaS buyouts is a) certain levels of consolidation, b) a deprioritization around the consumer experience, and c) a resulting business model that is disrupted when the consumer experience is fixed.
And when those factors are present, perhaps its time to build Metropolis for that industry with the explicit goal of further consolidation.
As for new business formation, when distribution is relatively easy to unlock, this strikes me as a far better place to start. Revenue growth at high capital efficiency is always better.
AI invokes an arms race. Vertical SaaS has in the past dealt arms to all sides. Perhaps its time to actually compete against industry incumbents and win.
The incumbent observation is that AI makes workers 30-50% more efficient.
Private equity feasting on service based industries should be evident of this.
Note: this is something I think is worth a far more extensive look at.
For the most part, the industries that look extremely attractive with AI innovations have incredibly low EBITDA/SDE multiples. The multiple stacking will alone create incredible returns.