Every SaaS product now has an "AI-powered" badge somewhere on the landing page. It's wedged between the hero section and the pricing table, right next to an illustration of a sparkle icon. But if you peel back the curtain on most of these features, you'll find the same thing: a thin wrapper around an API call, a prompt template, and a loading spinner. When every product has AI, no product has AI. The differentiator disappears. And the moat goes back to where it always was: product quality, distribution, and taste.

The commoditization curve

AI features have followed a remarkably compressed hype cycle. In early 2023, shipping anything that touched GPT-4 was enough to get press coverage and a spike in signups. By mid-2024, users expected it. By 2025, it was table stakes. McKinsey's research on AI in SaaS confirms this trajectory, noting that AI functionalities are rapidly "becoming table stakes in the business." What once justified a premium price now barely registers as a feature. According to High Alpha's 2024 SaaS Benchmarks Report, 56% of SaaS companies had launched or tested AI features within the past year, with the most common motivation being revenue expansion. But when more than half the market is doing the same thing, the expansion opportunity evaporates quickly. This is the classic commoditization curve, just running at AI speed. What took cloud computing a decade to normalize, AI features accomplished in under two years.

The GPT wrapper graveyard

The numbers are bleak. According to CB Insights, roughly 78% of AI startups launched in 2024 were essentially API wrappers, amounting to over 12,000 companies building on the same foundation models. The result was predictable: 966 U.S. startups closed in 2024, up 25.6% from 2023. In just the first quarter of 2024, 254 venture-backed startups filed for bankruptcy, a 60% jump year-over-year. AI wrapper startups suffered especially brutal retention numbers, with an average 65% churn rate within 90 days, nearly double the SaaS industry average. Users would try the product, realize they could get 80% of the value from ChatGPT directly, and leave. The pattern repeated across categories. AI writing assistants, meeting summarizers, tweet generators, code helpers: each wave of wrappers surged and then collapsed as the underlying models improved and platform providers absorbed the functionality. As one VC bluntly put it, "platform providers like OpenAI destroy entire wrapper categories overnight."

Why incumbents win the feature game

If adding AI is a checkbox, then the companies best positioned to check it are the ones who already have the users, the data, and the distribution. This is the incumbent advantage, and it's massive. Bain & Company's 2025 technology report frames this clearly: AI will disrupt SaaS in some cases by growing markets, and in others by commoditizing them. But in both scenarios, incumbents who move quickly hold a structural edge. They have proprietary training data from years of user activity. They have existing workflows that AI can enhance rather than replace. And they have the distribution channels to ship features to millions of users overnight. Consider how Notion, Figma, and Linear each added AI capabilities. None of them needed to build a new product. They layered intelligence into existing workflows where users already spent their time. The AI wasn't the product; it was an accelerant for the product. Software Equity Group's 2026 Annual Report found that AI-referenced deals comprised 72% of all SaaS transactions in 2025, a 12x increase since 2018. But crucially, the valuation premiums went to companies where AI deepened existing moats, not to those where AI was the moat.

The laundry detergent parallel

There's a useful analogy in consumer packaged goods. When the active ingredients in laundry detergent became standardized, every brand could clean your clothes equally well. The product was commoditized. So what determined which detergent you bought? Branding, shelf space, packaging, and distribution. AI features are heading the same way. When every project management tool can summarize tasks, every CRM can draft emails, and every analytics platform can generate insights, the AI itself stops being the reason anyone chooses your product. The reasons become the same ones they always were: does the product solve my problem well? Is it pleasant to use? Does it integrate with my stack? Do I trust the company behind it? The "AI-powered" badge becomes the SaaS equivalent of "new and improved" on a detergent bottle. It's expected. It's meaningless. And nobody makes a purchase decision based on it.

What actually differentiates

If AI features are commoditized, what separates the winners from the losers? A few things still matter enormously. Proprietary data. Companies sitting on unique, hard-to-replicate datasets can build AI features that competitors simply cannot copy by swapping in the same API. This is why vertical SaaS companies with deep domain data are often better positioned than horizontal tools. Your general-purpose AI writing assistant is competing with ChatGPT. A medical documentation tool trained on millions of clinical notes is not. Unique workflows. Cursor, the AI code editor, is technically a "GPT wrapper" in the loosest sense. But it reached $500 million in ARR with zero marketing spend because it built a deeply integrated development workflow that made AI feel like pair programming, not a chatbot in a sidebar. The wrapper was so good it stopped being a wrapper. Taste in UX. This is the least quantifiable advantage and perhaps the most durable. When the underlying technology is the same, the product that wins is the one that makes the right design choices. Which context to surface, when to be proactive versus passive, how to handle uncertainty. These decisions compound into experiences that feel fundamentally different even when the AI backbone is identical. Vertical depth. Guidewire's AI claims adjudication works because it operates in a narrow, high-stakes domain with specialized rules and regulations. A horizontal AI tool would need years to match that context. Going deep into a specific industry, with its specific data, workflows, and compliance requirements, creates a moat that breadth cannot replicate.

Stop leading with "AI-powered"

Here's the practical takeaway for founders. If your pitch deck leads with "AI-powered," you're already losing. Not because AI isn't valuable, but because it tells investors and customers nothing about your value. The strongest AI-era companies lead with the problem they solve. The AI is an implementation detail. Nobody cares that Cursor uses Claude under the hood. They care that it makes them write code faster. Nobody cares which model Notion AI calls. They care that their meeting notes get summarized without effort. This doesn't mean AI is unimportant. It means AI is infrastructure, not identity. The companies that treat it as a building material rather than a billboard are the ones building something durable. The startups that will survive the current shakeout share a pattern: they picked a hard problem in a specific domain, used AI to solve it in a way that gets better with usage, and never confused the technology for the product. The "AI-powered" badge had a good run. But the checkbox era is over. The next phase belongs to the companies that were always building something worth using, with or without the sparkle icon.

References