There's a lot of bold bullshit prognostication about how new generative artificial intelligence is going to kill the Software-as-a-Service (SaaS) business model because now anybody can build custom enterprise apps with a simple ChatGPT prompt. Nothing could be further from the truth.
AI is going to lead to more SaaS products that don't suck. But, along the way, AI is going to kill most SaaS startups -- because most early-stage SaaS startups suck.
I'll let SMB explain what I mean.
For those who don't get the joke, here's the definition of a Minimum Viable Product (MVP) from Wikipedia: "a version of a product with just enough features to be usable by early customers who can then provide feedback for future product development."
The cult of the Lean Startup that dominates Silicon Valley and most venture-backed SaaS companies produces almost nothing but MVP SaaS products that are half-assed on their best day by design. The idea is to get early customers to tolerate these half-baked offerings until the founding team learns enough to turn it into actual enterprise software. (Customers buy at the MVP stage because they assume they'll get permanently grandfathered into absurdly reduced early pricing in exchange for being guinea pigs. Also, sometimes MVPs sort of work.)
But today, thanks to AI, I can type a few sentences and get a half-assed prototype for free. I don't need to try -- let alone pay for -- an outside startup's SaaS MVP that's buggy, unreliable, and may not last long. I can get that kind of V1 crap in an afternoon of puttering, complete with mildly functional code I can hand off to a real engineer as a proof of concept.
No, these AI-generated prototypes won't be very good. Most SaaS startup MVPs aren't any good, either. The difference is I don't have to go to a SaaS startup to get a crappy MVP anymore. I can roll my own.
That doesn't mean startups are going away -- startups are doing great -- it means SaaS startups can no longer get away with barebones MVPs. The bar for an initial version of a product just got a lot higher, especially if you want someone to pay for it.
The absurd idea that ChatGPT can spit out a useful SaaS CRM anytime somebody prompts it with "build me a Hubspot clone" is just that: absurd. Anyone who has ever built commercial-grade SaaS software (and I've built a lot) knows that it's really hard and requires sweating a lot of complicated details that GenAI code-vomiters won't address (and that's before we discuss the issue of systems maintenance and required security compliance). As such, there will be plenty of market left for human-driven SaaS startups to claim.
And, because GenAI makes it easy to spin up prototypes, internal alpha releases are cheaper and faster than previously possible. It will be easier than ever to launch a SaaS startup thanks to GenAI. But it will be harder than ever to make a SaaS product that people will pay for.
GenAI will make it simple to create generic, barely useful SaaS apps. But hyper-focused SaaS apps that meet very specific needs at a high level of competency will be become much more valuable precisely because GenAI can't deliver that level of quality, and because maintaining that quality over time requires significant investment. Moreover, with actual SaaS, the costs of maintenance get spread over all customers, not incurred by a single customer running a bespoke in-house GenAI product. (In other words, build vs. buy economics still largely apply.)
The result will be an absolute explosion of niche, highly mature SaaS startups looking to claim extremely specialized market areas.
Which leads me to my final conclusion, one shared by my old colleague and current AI investor Rob May: venture-backed SaaS is probably dead.
GenAI makes rolling up custom migration tools super-cheap now, so SaaS switching costs are going to drop like crazy. Customer lock-in is going to be really hard to enforce. VCs like moats; GenAI is going to mass-produce bridges. As such, the only way to keep a customer on your product is for your product to actually be great. Being great is hard and expensive and may take a while. VCs aren't known for their patience, and they really hate customer churn.
Moreover, venture capital economics require that every company a VC invests in target a huge total addressable market (TAM), then set money on fire to try and claim a dominant position in the market before anyone else. The niche SaaS apps that AI can't create will have much smaller TAMs than VCs will tolerate. If you want to build a SaaS app in the future, be prepared to bootstrap.
In conclusion: The rise of code-generating artificial intelligence isn't going to destroy the SaaS business model -- just the SaaS business model as we've known it. The future of SaaS is a staggering variety of small, specialized, highly refined products that are ready for prime time at V1.
MVPs -- and VCs -- need not apply.
---
PS - Want to build a killer SaaS product before the GenAI market shift changes everything? I'm presently available for hire. You can grab my resume and generic cover letter here.
No comments:
Post a Comment