The idea that we’ll use chatbots for everything we use apps and software for a decade from now is frankly absurd.
Maybe chatbots will be more useful than they’ve ever been, but replacing all software with something like ChatGPT seems far-fetched to me. And in that sense, the SaaSpacolypse has been overblown.
What will change is the business model behind the software we use. I think that’s undeniable, at this point.
We’ve seen business model shifts in software, and they aren’t often smooth, but they can be moments of opportunity for investors.
My childhood was spent putting floppy disks and then CDs into computers to install everything from games to antivirus software. Sometime in the late 2000s, that was replaced by downloadable software that ultimately led to the SaaS revolution. Instead of paying for a disk, customers started paying for a “seat” of software.
Companies like $Adobe(ADBE)$ went through a difficult transition from one-time sales to the SaaS model (2011 to 2014), but when it fully transitioned, growth was a simple formula of adding seats and slowly raising prices.
But that may not be the best formula going forward. Seats don’t matter as much as they used to in a world of AI. So, what will win? More on that in a moment.
Jobs as a Service
What if SaaS is the past for software?
What if the business model of the future is JaaS (Jobs as a Service)? (if you have a better naming idea, I’m all ears!)
I’m going to use Autodesk as an example, but what I discuss could be true for dozens of other software companies.
Today, an engineer starts with a blank screen when designing a product or a manufacturing line and spends weeks or months building it up. Parts are drawn in 2D and “extruded” and then holes, edges, and other modifications are added. There are design reviews, and eventually fluid dynamics, structural analysis, and manufacturability tests (ex can the part be injection molded?) may need to be done, resulting in revisions that need to be tested with specific software (like Ansys).
The process has gotten easier over time. I began drawing on a 2D screen with Autodesk and eventually moved to Inventor and Solidworks in college, which was more than 20 years ago. Since then, the field has gotten much more advanced.
But there’s still not a request like, “I need an airplane wing that can withstand X loads, build an optimized wing for me.”
Maybe AI can do that? The systems would need to be designed to allow for that kind of request, but (in my mind) design and engineering are more like AI for coding than AI for writing. There’s a request to do something, and then there are constraints that can be tested and iterated on, which is what AI seems to do well.
I coding, that’s “does the software run?”
In the design areas I’m talking about, it’s “does the part do what I asked?”
Let’s say this future world exists. What is the business model?
A per-seat license probably doesn’t make sense, even at the high price point these tools go for today. Even at $3,000+ per seat, a suite of software from Autodesk probably understates how valuable the software is. But if it’s doing the job for the designer and 10xing their productivity, that’s an order of magnitude more valuable.
A price per job probably makes more sense.
But even that will need to have constraints.
Maybe it goes like this:
Prompt: I want this part. I want to make this part. Here are the constraints. Build a proposal.
Answer: I can optimize your part for weight, strength, and manufacturability for $150 of tokens, provide a working solution for $50, or make a minimum viable solution for $10.
Depending on where you are in the design process, that would be a great solution.
Imagine the same for architects.
Prompt: Design 5 home options for a prospective client on this property with 5 bedrooms, 3 baths, 3-car garage, and room for 2 adults, 3 kids, and 2 dogs.
Answer: I can provide full blueprints and engineering specs for $2,000 or a proposal package for $100.
In theory, a single product designer, architect, engineer, etc., could have a dozen “agents” working for them at once.
Productivity goes through the roof!
The opportunity in this scenario is not about data or how ingrained a piece of software is in a business (like ServiceNow or Salesforce), but rather about how powerful the AI can be within the software’s constraints.
If you think about the rocketships that are Cursor or Anthropic, they’re built on making software engineers 10x more productive. They have the systems to make that happen and test what they’re doing, allowing agents to run for hours.
If the same is true in other parts of software, it will be about what programs can become productive workers?
Something to think about, and it’s an area I’ll be exploring because the JaaS model may result in an explosion in revenue for these software companies. And this time, the transition to a more sustainable business model may be less painful than it was going from selling disks to the SaaS model we know today.
😍 Been eyeing Tiger merch but short on Tiger Coins? Now's your chance.
🎁 We’ve selected 4 high-demand items across practial, lifestyle, and learning, now with a lower redemption threshold!
Hot Merch Returns · Up to 43% Off
Comments