AI is officially beyond hype. This is a full-blown evolutionary event we’re experiencing, and it might mean the death of SaaS (software-as-a-service products).¹

That's what Microsoft’s CEO Satya Nadella seems to think. Here’s what I think about why SaaS is dying, and what other types of products will thrive in its place.

How SaaS is Dying

Why might SaaS be dead? First, backstory.

It’s now obvious that AI agents have the potential to augment (if not completely replace) a vast number of internet tasks, especially in areas like content creation, data processing, search/discovery, and automation.

For the past 5-10 years many of these tasks have been addressed by SaaS products — which by definition are cloud-hosted, subscription-based web applications. Examples:

• Content Creation: Grammarly ($23/mo), Canva ($13/mo)

• Data Processing: Tableau ($35/mo), PowerBI ($10/mo)

• Search/Discovery: Crunchbase ($49/mo), Statista ($39/mo), Ahrefs ($99/mo)

• Automation: Zapier ($20/mo), Asana ($14/mo)

• Communication: Mailchimp ($17/mo), Notion ($10/user/month)

• Transactions/Contracts: Stripe ($-), DocuSign ($25/mo), TypeForm ($25/mo)

These examples are the best-of-the-best SaaS products — they represent the 1%. But even these are already being eclipsed by today’s standard chat models. A perfect example is Grammarly — I used to pay for a monthly subscription, now I use Claude / ChatGPT instead (and get better results!)

Mind you, that’s what’s already happening to the best of the best. But what about the other 99% of SaaS products? Many are simple, one-feature products, early-stage startups like my own Uptrends.ai, and many don’t stand a chance.

In short, 99% of today’s SaaS products will either die or adapt. Many already have died (ex: TinyURL, Removebg), and many more are adapting to incorporate AI as quickly as possible.

Importantly, SaaS’s AI problem isn’t one dimensional. It’s two-fold. Not only do AI agents produce better results than incumbent SaaS products in many of the aforementioned areas, but AI has also made it incredibly easy for people to spin up new competitors to existing SaaS products and compete on price.

You can clone Notion with AI tools like Bolt in minutes, like this guy did. Meaning not only has AI made it harder for SaaS products to compete on quality, but increasingly on price as well.

What to Build if Not SaaS

So if you’re a SaaS person, or an aspiring founder looking to start a new internet business in 2025, or an investor looking to reposition their portfolio as AI continues to advance and SaaS dies, the million dollar question: what should startups build in the second half of the 2020s if not SaaS?

Rather than building SaaS — where everything becomes a database ‘wrapper’ with a fancy UI/UX competing on price — I’d focus on other areas of the puzzle. Here are the 5 most compelling business types in this world of AI-eats-SaaS:

1. AI Native Companies

This is the most obvious answer… in a world dominated by AI, you could just build AI. This means specifically incorporating agentic / AI workflows into your product feature set, ideally at the core, rather than as an afterthought. Many existing SaaS companies (Grammarly, Notion) are trying to evolve towards incorporating AI — stapling on generative chat functionality to their existing feature sets. So if you can start from scratch with it built into your business at the fundamental level (from data to pricing) you have an advantage over the incumbents.

2. Infrastructure Companies

Of course as AI becomes more ubiquitious, there will inevitably be a need for more infrastructure to support it — much like we’ve seen with previous tech revolutions like the internet, cellular, etc. This is where most of the money in AI has been made so far over the past 5 years; building the platforms and infrastructure on top of which the AI-native companies will be able to grow. AI model hosting, specialized computing hardware, data pipelines, and edge computing solutions (Nvidia, AWS, Databricks, Snowflake, Huggingface, etc.) are worth paying attention to, but this space has high startup costs and white space is sparce.

3. Data Companies

Not sure if this is a hot take or old-man take, but one of the unsung heroes of the AI revolution will be the data companies. Everyone loves a good, well-spoken LLM-generated output, but the real magic is in the training data (some argue that’s ALL of the magic). It’s become cliche to say that data is the new oil, but it’s true.

Looking forward companies that aggregate, clean, and structure high-quality data – especially in specific industries and verticals – will be essential. The key is (say it with me) proprietary datasets. The startup costs are generally lower, and for many existing SaaS companies, this may prove to be a great alternative revenue stream to subscriptions as AI continues to evolve.

4. Content Creators

The first 3 here cover the technology and how it will change. But what about the rest of the pie? As AI and software continue to eat the world, one thing becomes exponentially more powerful: content. It’s the vehicle by which we communicate (and sell / market / distribute our tech products), and it takes up an increasing portion of our waking time spent.

My advice to any person young or old (which I am not qualified to give) is to build a content library. Build a personal brand. Become a freaking micro-influencer in whatever domain you have a sliver of expertise in. It’s the most powerful tool of the future — an owned audience. As Naval famously tweeted: “Be a creator and you won’t have to worry about jobs, careers, and AI.”

5. Service Providers

Last but not least is the service space. Good ol’ fashioned hand-made ingenuity, for all the people in the back. Non-tech jobs will grow at a premium over the coming years, and tech-enabled services will too.

Anything that continues to require a human’s touch will maintain an edge in the shrinking world outside of AI. This extends not only to consulting, advising, fractional CXO, whatever it is you can do to get paid for having expertise in. It also extends to landscapers, interior designers, sidewalk shovelers, and barbers.

The long story is now short, SaaS as we know it is evolving — maybe not dying, but surely changing and becoming thinner.

If anything dies it’ll be single-feature micro-SaaS products. Increased AI-generated competition will either drive their margins to zero, or larger incumbents will add similar features to more robust products making micro-SaaS’s like them obsolete.

In this world, it seems better to focus on the peripheral pieces outside of subscription-based cloud-hosted web interfaces wrapped over a simple CRUD database. You should focus on:

• The infrastructure that supports AI

• The data that trains AI

• The content that distributes AI

• The services that enable AI

• Or the AI itself

Curious to hear your thoughts. This one was very close to home for me, and I imagine I’ll be looking back on this article a few times down the road to see how my predictions maintain temperature. In the meantime, enjoy, and cheers to 2025 and the crazy future beyond.

Cheers,
Ramsey

1  SaaS = software as a service, differentiated from traditional ‘software’ by the combination of cloud delivery and subscription pricing.