What are the most and least promising categories to invest in, within seed-stage AI SaaS?

A primary strategic division has emerged in AI SaaS investing, separating the infrastructure and application layers. The infrastructure layer, encompassing data platforms and foundational models, is widely considered the most durable and essential component of the new AI stack [1, 22]. Analysts view companies providing this fundamental "plumbing" as direct beneficiaries of the AI boom, offering a more defensible investment thesis [5, 30]. In stark contrast, the application layer is characterized by significant uncertainty, as investors question the long-term terminal value of incumbent SaaS companies and rotate capital towards foundational players [1, 9, 13, 20]. While this dynamic creates an existential threat for some application providers, it simultaneously opens opportunities for entirely new, native AI categories that large corporations are structurally too risk-averse to build [4, 23].

The most promising opportunity for seed-stage investment appears to be vertical AI SaaS, particularly in industries with complex, unstructured data and a lack of dominant software incumbents, such as legal, healthcare, and finance [4, 18, 21]. These companies are predicted to become at least **10 times larger** than their traditional SaaS counterparts because they can capture value from operational budgets, not just limited software spending . Defensibility for these startups is derived from building finished products by layering AI over proprietary, "walled garden" datasets, creating a significant competitive moat that is difficult to replicate [3, 14]. This targeted approach contrasts sharply with broad horizontal applications like office productivity co-pilots, which enterprises are still experimenting with and which have yet to demonstrate clear ROI .

Conversely, the least promising investments are generic, thinly-wrapped "AI companies," with one venture capitalist predicting that seed funds investing in this category will lose an **impossibly large amount of money** . Significant skepticism surrounds the current investment cycle, with some experts arguing the odds of generating a 100x return are now "really, really low" and comparing the environment to the first wave of the dot-com bubble . This perspective suggests the most valuable companies may not be founded for another 2-5 years, following a market correction . This caution tempers the "Sasspocalypse" narrative, which some view as overblown; incumbents with deep enterprise integrations are proving resilient, with 75% of public SaaS companies having successfully raised prices since ChatGPT's release [3, 16, 19]. This indicates investors must perform nuanced analysis to differentiate between SaaS companies genuinely threatened by AI and entrenched systems of record that are enhanced by it .