AI has not killed SaaS. It has killed bad SaaS

For a SaaS founder like me, the “SaaSpocalypse” is not just a market story. It is a question you start to ask yourself daily: if AI can generate software faster, automate workflows, and sit inside the tools customers already use, why should anyone keep paying for your product?
I think it is the right question. I just think many people answer it wrong. SaaS is still worth building, but only if the product is useful enough, trusted enough and focused enough to earn its place.

I am a Ukrainian founder living and building in London, and I co-founded Bluedot, an AI note-taking and meeting assistant company. We are building in one of the most competitive categories in software. Google Meet has Gemini’s “Take notes for me”. Otter is well known. Granola has built a strong product experience. Fireflies and Fathom are also everywhere in the category. Every tool is promising notes, summaries, CRM updates, productivity and automation. So I do not have the luxury of pretending this market is easy. Features can look similar. Customers can try alternatives quickly. The biggest platforms already own the calendar, the meeting room and the document where the notes might live.

But Bluedot has reached more than $4m in revenue in just a few years. I would love to say this is because we are just very cool and good, but in reality we were also lucky to end up in a market that started growing fast. The global AI note-taking market is estimated at $740.4m in 2026 and forecast to reach $3.48bn by 2035, according to Precedence Research. A growing market does not make company-building easier. It usually means more competitors, more noise and more customers trying five or ten tools before deciding which one they trust. What our experience shows is that businesses are still buying software. They are just much less patient with software they do not trust, do not understand or do not need enough.

That is what I think the SaaSpocalypse really means. Reports said roughly $285bn was wiped from SaaS valuations in 48 hours in 2026, as investors priced in the idea that AI agents could replace whole categories of subscription software. At the same time, Gartner expects global software spending to reach $1.44tn in 2026, up 15.1%. The market is still there. It is just harder to defend average software.

Good SaaS owns a workflow, not just a feature
AI makes prototypes easier. It makes feature-building faster. It lets smaller teams look bigger than they are. That is good for founders, but it also removes some protection. If your company is only a feature, someone else can probably build something similar. This is where big platforms have an advantage. If Google adds AI notes inside Meet, customers already understand why they use Google Meet. If Microsoft adds meeting intelligence into Teams, customers already have Teams open all day. They do not need to explain the workflow because they already sit inside it.
For a startup, the risk is different. If you add ten disconnected AI features, customers may stop understanding what you are actually for. This is what I think bad SaaS usually is. It is not only bad design or bad pricing. It is lack of focus.

At Bluedot, we have different types of users, but the core use case is clear: note-taking and meeting intelligence. We try to add features that make that use case better, not just features that are possible because AI exists. This sounds simple, but it is easy to forget when every competitor is launching something and every investor wants to hear how you are using AI. For us, the problem is not “meeting notes”. The problem is that important information disappears inside conversations. Follow-ups get missed. CRMs are not updated properly. People leave meetings trying to remember what actually happened. When that problem is solved well, the product becomes more than a transcript. It becomes part of how a team works. That is why I think the best SaaS companies own workflows, not features. A feature can be copied. A workflow that customers depend on takes longer to build, but it also becomes harder to remove.

A crowded market makes trust more valuable
People sometimes think a crowded market is only bad. I do not fully agree. A crowded market means the problem is real. If there were no other AI note-taking tools, maybe that would mean nobody cared enough about the problem. The problem with a crowded market is choice. When there are too many products, people do not want to evaluate everything from zero. They ask friends, colleagues, founders and investors. They look for the product someone they trust already uses. That has been our experience. People who come from ads often browse. People who come from recommendations often buy. They already have some trust because someone else reduced the risk for them. This is also why SaaS still has a service element. People forget that SaaS means software as a service. The second part matters more now, not less. At Bluedot, we have won deals not only because of the product, but because customers felt there were real people behind it. We try to speak with customers, understand their goals, be proactive and help them succeed.

In B2B, people are buying confidence as much as functionality. They want to know that when something breaks, someone will care. They want to know that the company understands their workflow. They want to know that the product will still be there next month and that the people behind it are serious. I think about this like choosing a coffee shop. I was speaking with my co-founder Ruslan, who was in Turkey, and he told me about a street with ten coffee shops. All of them looked good from the outside, and he tried a different one each day, but eventually he kept going back to the one that made the best coffee, even though the interior was not the best.
Software is similar. Marketing can get someone to try you once, but the product has to make them come back.

AI makes building faster, but reliability harder
AI can make development faster, but founders should be careful not to confuse speed with quality. I was speaking with our CTO about this recently. His point was that when you build something yourself, you have to go deep into the problem. You think about the possible solutions, the edge cases and what can break. When you rely too much on AI-generated code, it is easier to skip some of that thinking. The product can look good on the surface, but the understanding underneath may be weaker. For a prototype, that might be fine. For a product a business depends on, it is dangerous. This is not just about coding. It is about how founders think. If AI makes it easy to build a feature, you still need to ask whether the feature should exist. Does it make the workflow better? Does it make the customer more confident? Does it reduce work, or does it just create another button? SaaSBench, a 2026 benchmark for enterprise SaaS engineering agents, found that over 95% of task failures happened before agents even reached deep business logic. The main bottleneck was not writing isolated code, but configuring and integrating multi-component systems. That is why “AI can build software” and “AI can build a reliable company” are very different things. Launching is easier now. Reliability is still hard. Review, testing, security, infrastructure, support and customer trust still take time. AI can help with many of these things, but it does not remove the need for discipline.

What customers actually ask before they buy
The biggest questions we get from customers are often not about the most impressive AI feature. They ask where their data is stored. They ask whether we train models on their conversations. They ask how secure the product is, what happens if something breaks, and how it integrates with the tools they already use. This is especially true in the UK, where many businesses are interested in AI but still cautious about using it in real workflows. YouGov found that among UK SMEs using software solutions, 65% cite reliability and accuracy as a barrier to AI adoption, while 51% cite data security and privacy. This matches what we hear from customers. The first question is rarely “how clever is the AI?” It is usually whether they can trust this with their data, their meetings and their customers.

Most customers do not describe the value to us as “AI”. They talk about the thing they no longer have to worry about. In our case, people want to know that a meeting was captured, important details were not lost, the CRM can be updated and they can relax a little bit during the conversation instead of trying to write everything down. Yes, that saves time. But the real value is confidence. Good AI software does not just automate work. It removes anxiety. That is why security and reliability have to be treated as product features. SOC 2, ISO, penetration testing, clear policies, proper data handling and someone responsible for security are not enterprise checkboxes. They are part of why a customer can trust you enough to use the product in a real workflow. My advice to SaaS founders is simple: spend less, stay focused, and survive long enough for trust to compound. The winners will be the companies customers would miss if they disappeared. Good SaaS is still worth building. It just has to be good enough to deserve its place.