The Nine Archetypes of AI-First SaaS Products
Plus labor shortage data and a business breakdown
The Big Picture
Data points and market trends
In a recent podcast I listened to, a financial analyst explained that by the end of this decade, we could be looking at a global labor shortage of around 40 million workers, which translates to $3 trillion in wages that won't be spent.
And he pointed out that businesses will naturally channel this budget in hardware and software solutions to tackle this labor shortage issue.
And it’s not just sectors like manufacturing or healthcare (in which labor shortage is severe) that will benefit from it; white-collar industries are also experiencing it.
For instance, the financial and business services sector might face a $1.3 trillion revenue shortfall each year due to a projected shortage of 10.7 million workers by 2030. Same with the cybersecurity industry that could see a shortfall of nearly 4 million workers by the same time. Rinse and repeat with many other industries.
This essentially means that we could see several trillion dollars being funneled into both new and existing software categories by the end of the decade.
While we might be experiencing a SaaS recession right now, the current labor shortage coupled with the rise of AI could potentially lead to an interesting rebound in the next couple of years (and many opportunities for entrepreneurs).
Article of the week
Research and thoughts
The Nine Archetypes of AI-First SaaS Products
I analyzed around 130 AI-first SaaS products and realized that we could categorize them into nine major “product archetypes”.
In no particular order:
The Magical All-in-one software
Description: I currently see many founders building all-in-one software products from the ground up with AI at their core. All-in-one software (what I call swiss army knife here) is a class of products characterized by offering tens of features. Today, in the majority of the software categories and verticals, the SaaS incumbents are all-in-one software companies that have often evolved from point solutions. And many founders want to disrupt them with AI powered versions of them. The thesis behind is that it will be very difficult for existing all-in-one software products to reinvent themselves with AI and that you need to start from a blank slate to do it properly.
Examples: Nory is reinventing ERP for restaurants, Light is doing it for accounting.
Autonomous agents
Description: Autonomous agents are AI-driven software capable of performing tasks and making decisions independently without human intervention. The thesis behind this approach is that AI will increasingly automate tasks that were not possible with classic automation because GenAI can better understand the specific context of a situation and ”improvise” a solution by itself.
Examples: Autify is an autonomous AI agent for software quality assurance, Radiant autonomously triages and investigates security alerts, Superagent for web search…
Intelligent assistants
Description: AI assistants are AI-powered tools that help people perform tasks more efficiently by providing guidance, support, and recommendations without taking over the task completely. It’s basically the Co-pilot model. The thesis behind this model is that for many tasks we are not yet at the stage where AI can autonomously complete them, but it can help humans complete them much faster.
Examples: Respaid helps businesses collect unpaid invoices, Kelvin helps construction companies fill energy renovation related paperwork…
Opportunity discovery software
Description: Opportunity Discovery AI software helps individuals and businesses uncover new revenue or cost-saving opportunities by automating the processes of identification, analysis, and execution through advanced AI capabilities. I wrote a more detailed post about this approach here. The thesis behind this model is that since GenAI is good at understanding and searching large corpus of data (whether text, audio or video based), it can help businesses discover opportunities that an average human could find.
Examples: Tengo enables companies to discover the public tenders they can answer to. Metris helps real estate companies develop the solar potential of their portfolio. Camion helps charge point operators find new places where to install EV charging stations. JobRight for job search…
Competitive practice AI software
Description: Competitive Practice AI software enables businesses to quickly replicate successful strategies by automating the adoption of proven competitive practices, leveraging AI's capabilities in analysis, discovery, and recommendation. I wrote a more detailed post about this model here. The thesis behind this approach is that GenAI is very good at recognizing patterns and now automates the copycating process.
Examples: Dora generates landing pages with the latest design trends, AdCreative for ads…
Content and data generation
Description: Content and data generation AI software enables businesses to create content tailored to their industry needs. The first wave of content generation startups focused on low-hanging fruit cases (generating blog posts, profile pictures, etc.), but the latest wave is addressing more specific/vertical use cases (see the examples below). The thesis behind this approach is that GenAI is very good at generating content that humans would normally create. This applies to a wide range of content, whether it's text, visual, or audio-based.
Examples: Sloyd generates 3D models for gaming companies), Veeton generates images for the fashion industry, Makepodcast generates audio podcasts, Fairgen generates synthetic data for market research…
Intelligent search
Description: LLMs are really good at understanding large corpus of unstructured data, whether it’s text, video, or audio-based. And a first use case is to leverage this capability to enable businesses to search through these corpus. The thesis behind this approach is that businesses are sitting on troves of unstructured data that they can now search thanks to AI.
Examples: Realm or LangDock for internal knowledge, Review for legal documents, Daydream for ecommerce…
Data extraction and meaning
Description: Another interesting application of AI's ability to understand unstructured data is to extract meaningful insights from it. I see an increasing number of startups using AI to derive meaning and context from data to complete subsequent tasks. The thesis behind it is that large language models (LLMs) will power the next generation automation software, enhancing it with better functionalities and more use cases.
Examples: Pixacare for healthcare (don’t open this page if you’re having lunch), SewerAI for sewer inspection, Sketch for counting materials from building drawings…
Digital twin software.
Description: The digital twin model refers to the creation of a virtual replica of a physical entity, such as a product, system, an entire ecosystem or even a person. This digital replica allows startups to simulate, analyze, and optimize real-world processes and products in a virtual environment.
Examples: Tibo Energy or Fever for energy networks, j4Energy for industrial sites, Pelikan for EV fleets…
Business Breakdown
Interesting startups and products
I recently stumbled upon SewerAI which operates in the sewer infrastructure market and provides AI-powered inspection and assessment solutions for underground infrastructure to utilities, engineers, and contractors. They just raised $15M and I think it’s a great example of how AI can improve the productivity of people working in the “real world”.
What I find interesting:
TIC (Testing, Inspection, Certification) is a very deep market. I had no idea, but the broader inspection market employs millions of people worldwide whose job is to inspect and check infrastructure or equipment.
AI can improve productivity and help with labor shortages. Inspection-related tasks are very manual-intensive, and there’s unsurprisingly a labor shortage there too. AI's capacity to understand images and video to extract data and automate tasks is a really good fit for this industry."
Come for the data and stay for the tool. As usual with data extraction tools, it’s not only the data extraction part that is interesting but also the software that can be built around it.
SaaS Framework
I would categorize SewerAI the following way according to our framework: