Updated: May 23
If you can code and have a good idea, you can release an MVP, raise finance, and start a business. Cloud-based infrastructure has become simple to deploy and manage, and you can connect (using APIs) with almost any other platform, including developer-friendly payment platforms, such as Stripe. Software cannot be protected (copyright), so your best defence is to scale – quickly. Your job is to go to market, test, learn, and repeat until you find (or not, as the case may be) product-market fit. Then raise more finance and scale by investing in sales and marketing. Billions of investment via 10-year venture capital funds, or tax-driven investment vehicles in the UK, have been poured into supporting thousands of such 'software-as-a-service' (SaaS) companies.
Is the party over the the SaaS business model?
The problem is that it looks like the party is over for the SaaS model and for a significant amount of the money already invested in funds with several years left to run, as technology experiences fundamental and increasingly exponential change. The valuations of the majority of SaaS investments are traveling in the wrong direction. Last week, I wanted to bulk transcribe 70+ videos to ingest into a GPT (general pre-trained transformer) that could then respond to questions via PiersBot, which is a chatbot I have launched that is being trained on all of my content, including my 76-lesson course for entrepreneurs.
I searched for solutions and found several SaaS companies offering the service subject to a subscription. Instead, as I like to know how stuff works, I registered on GitHub, installed a Python editor, watched two YouTube videos, and built my own solution using OpenAI’s powerful Whisper, a general-purpose speech-recognition model. All I had to do was stand up a website integrated into Stripe, and I could launch my own SaaS startup that would have a window of opportunity - of about a week.
The thin layer for SaaS
My own example demonstrates that increasingly SaaS companies are a wafer-thin layer of code over large-scale infrastructure platforms, large language models (LLMs), General Purpose Transformers (GPTs), and now AutoGPTs. With the launch of Amazon Bedrock and AWS Code Whisperer, which turns natural language into code, even the remaining thin layer of code can be written by anyone within a year, although it won’t be 'code'; only the input (natural language) and output (the outcome sought, such as a booked travel itinerary) will be visible. Few know or even care about how a microwave works. We just like the output - reheated pizza.
AI and natural language interaction
The thin layer that SaaS businesses increasingly inhabit is being eroded by the month. As natural language interaction with technology and AI platforms (such as GPTs) with specialist training (e.g., travel, finance, transcription) begin to communicate with each other to perform tasks, the need for SaaS to connect humans with services is being eaten.
AI platform collabroation (AutoGPTs)
Take your favorite takeaway delivery company, or collection of travel, car hire, and hotel booking websites. AutoGPTs are still in their early days, but they can perform all of these tasks and keep you posted and optimize them over time should anything change, such as room rates before you leave.
Who can creating competitive advantage and build a moat?
There are opportunities to create moats, and getting there first is still an expensive strategy that could work. In an exponential world, getting ahead means staying ahead as catching up becomes more difficult. However, in most cases, a sustainable competitive advantage will require huge scale and access to large amounts of capital. Here are some ways businesses can create competitive advantage:
Deep Tech: AI has not yet reached the level of expertise needed to fully grasp the complexities of deep tech, although it will get there, and super-intelligence will render most human brainpower redundant.
Large-scale Processing: AI models often require significant computational power to function effectively. This opportunity already belongs to Microsoft, Amazon, and Google and could bring them back from being 'post-growth.'
Data Ownership: Ownership of large and proprietary datasets on which AI can be trained is an advantage, although this will also be transient as they can be bought.
First Mover: Any business that can stay ahead in an AI arms race may be able to stay ahead.
Creativity: Combining software with creativity and services that can't be automated, although such business models may be less scalable.
The difficulty of picking winners and losers
The winners from Web 1.0 were not the obvious ones in the late 1990s. Although the opportunities presented by AI are enormous and could change the course of humanity, selecting the winners and losers during the current period of hype and noise is almost impossible. A SaaS business started today could be automated out of a revenue opportunity next month.
Lets not forget that LLMs and GPTs could be the betamax (lost to VHS video format - for non GenX readers, VHS was a format before DVD which was the last physical format before streaming) of AI.
The diminishing opportunity for SaaS businesses
The layer in which SaaS businesses can operate will eventually be unable to support life.
Marc Andreesen famously wrote in a 2011 article that software will eat the world.
Software (AI) is going to eat sofware (SaaS) first.