AI or Not to AI? The Dilemma Every SaaS Founder Faces
We're living in the AI era. Many founders today feel pressured to build an AI-first product because that's where the excitement is. The trend is understandable, but that does not mean your SaaS product has to be AI-first to be marketable and profitable.
Journey with me!
Avoiding the AI-Powered Product Charade as a way to win the market
Every week, a new AI startup launches. Every product seems to be adding an AI copilot. Investors ask about AI. Some customers (maybe) ask about AI. Even founders feel pressured to mention AI somewhere (as a source of differentiation) in their product description, telling users it is a feature that is integrated into the product without caring if users really want it, let alone use it.
SaaS doesn't win the market simply because you build it as an AI-first product. In fact, I believe that when building a SaaS product, founders should be cautious of ending up with a costly and complex product all in the name of building a product that is AI-powered, since that is the new way of things.
Making a product AI-First means the architecture roughly follows something like this.
User → Product accepts a problem (LLM Call) →solve it (LLM Call)→Present the solution to the User (LLM Call)
Every single step is routed through an LLM call. It feels modern. It feels "AI-native." It's also, in a lot of cases, a worse product than the boring alternative — and more expensive to run, forever.
I call this the "Wrong Product Thinking Approach": starting with "we have AI, where can we use it" This makes it hard to distribute because the moat is weak, if non-existent, because AI is more of a technology and not the solution that users pay for. My argument here is that users don't care what's under the hood. They don't know, and don't care, whether their answer came from GPT-4o, Claude, or a hardcoded function block that never made an API call in its life. They just want to see the result. In fact, they care more about whether it's fast, correct, and cheap enough that you're not passing the cost back.
You, on the other hand, know exactly what's running behind that button. And you know exactly how much OPEX you're incurring for every LLM API call per user, per task or per session. The same goes for how much you are saving every time a problem gets solved without an inference call.
Building a Profitable SaaS while going with AI Flow
On average, plenty of profitable SaaS products don't really need AI to sell well. Think about products like inventory management systems, pharmacy management software, restaurant POS systems, and school management systems, among others. The core value of products is not making them AI-first for users. Rather, they come from making business operations smoother, faster, and more reliable, and that may not necessarily require the integration of AI to be achievable
Don't get me wrong: AI actually adds value, and I am not advocating against the use of AI or its integration into SaaS products. Far from it. I am making a case for building a profitable SaaS that people use and conveniently pay for. Instead of thinking/asking: "How Do I Add AI?" when working on an idea, you should be asking questions like:
- Is AI needed to solve this problem?
- Does AI create enough value to justify the added complexity?
- How does AI affect the cost of the product for the user?
- What is the overall business gain if AI is to be added to this product?
Questions like these need serious consideration because the real value of a SaaS should not be rooted in its AI-First Nature. Long before large language models became mainstream, as we have been seeing for some time now, SaaS products were already solving real business problems. Their real value is largely in the way they help businesses and individual users achieve one or all of the following:
- Save time
- Reduce human error
- Automate repetitive work
- Centralize information
- Standardize workflows
- Make everyday operations easier
None of those outcomes requires the SaaS product to be AI-first. They require solving a painful problem exceptionally well. Whether those outcomes come from AI or carefully designed automation is often secondary. AI is great and will continue to be. But there's a difference between using AI as a feature and making AI your entire product strategy. AI use and integration become incredibly valuable when it improves an already useful workflow. SaaS builders who understand this are those who will continue to build and distribute what people use profitably.
Until next time!
Gambatte!
