Over the past few months, public software stocks have experienced sharp corrections: the BVP Emerging Cloud Index (EMCLOUD) down 20-30%, with individual names dropping 20%+ on frontier model announcements from Anthropic and OpenAI. Investor communities are asking: “Has AI killed SaaS?”
Our view is that AI is accelerating a bifurcation that was already underway. Undifferentiated software products without defensible data, distribution, or domain expertise face compression. Meanwhile, specialized applications with tangible moats are using AI to compound their advantages. The companies that fail won't fail because AI exists; they'll fail because their solution wasn’t durable enough in the first place.
AI capabilities such as data synthesis, pattern recognition, and code generation deliver real productivity gains. But enterprise deployments, particularly in regulated industries, still require accuracy guarantees, workflow integration, and institutional trust that general-purpose platforms simply can't provide. This gap creates the opportunity.
At Laconia, we focus not on the capital-intensive infrastructure-building portion of AI (i.e. frontier AI labs) but instead on the capital-efficient, scalable, specialized software application and platform opportunities. These are the investments where we can repeatably identify high upside opportunities with structural entry points that fit our seed-focused strategy. As we enter 2026, we are reflecting on what opportunities emerge as a result of both the capabilities and the limitations of AI. What's different, what's the same, and where are the diamonds in the rough?
What's Different
Faster time-to-product expands and crowds the opportunity set. Building an MVP is cheaper and faster, making previously unviable niche products economically feasible. This increases total experimentation and, eventually, venture-scale opportunities. But it also means:
Supercharged competition: We see dozens of near-identical businesses, often with $100k-$800k in ARR, across a few to a few dozen customers, within a few months of incorporation. With this landscape, defensible differentiation and strategic vision matter much more than early traction or market signal.
"Execution plays" are dead, and companies need structural defensibility. There used to be a whole category of businesses with no long-term moat where it was not unreasonable for investors to bet that the founders would win simply through superior operational execution — better sales, faster shipping, stronger fundraising. AI makes this investment proposition untenable as competition for low-hanging fruit is lethal. Without a clear structural advantage that competitors can’t replicate – proprietary data sources, exclusive distribution partnerships, deep domain expertise in regulated industries, or multi-year technical moats – software companies face a race to the bottom on price. Pure operational execution remains necessary but insufficient. AI has made the distinction between "doing the same thing better" and "doing something others can't do" the line between viable and non-viable seed investments. Here are a few examples from our portfolio:
Ocrolus (Fund II): Human-in-the-loop verification combined with machine learning delivers near-perfect accuracy in financial document analysis. This operational moat can't be replicated with prompts alone. Interestingly enough, the “human-in-the-loop” element was one of the most frequent reasons that investors passed on Ocrolus in the early days, and it has proven to be one of its biggest advantages.
Messium (Fund III): Messium combines exclusive hyperspectral satellite imagery with multi-season ground-truth soil data and strong channel relationships. The data advantage and distribution are slow to replicate.
Every startup’s competitive set now includes hyperscalers. The evergreen VC cliché of "What if Google (or Microsoft, or OpenAI, or Anthropic) builds this?" is real now. Software startups are not just up against non-technical incumbents like publishers or taxi companies anymore; they are up against the most sophisticated and well-capitalized technology businesses in history. One archetype that works well is a vertical-specific product that is 1) specialized enough that larger tech platforms can't match relevant market expertise and 2) up against legacy incumbents that lack the technical capability to deliver a compelling product experience. Add an underrated market size and a go-to-market Trojan horse wedge that reduces existing vendor lock-in, and you might have a winner.
Business models are shifting. Pure subscription revenue is declining; usage-based, metered, or workflow-integrated pricing better aligns with AI value delivery.
"Software+" models are viable. When software was a bottleneck to product development, there was only so much that a company could tackle at once. AI lowers the software-building constraint, making targeted combinations of software + hardware or human-in-the-loop economically viable, assuming the unit economics trend toward software margins at scale.
What Hasn't Changed
Value accrues to solutions, not tech stacks. Lasting adoption requires clear ROI, workflow-native design, deep customer understanding, and product simplicity. If software businesses fail, it won't be because AI now exists – it'll be because they didn't get the experience right.
Teams drive outcomes. Judgment, speed of learning, rigorous processes, and leadership quality remain the key attributes that we seek.
Unit economics are non-negotiable. Gross margins, payback periods, and retention matter as much as ever.
Capital strategy shapes outcomes. Entry valuation, ownership, and fundraising trajectory must match not only the company’s operating needs but also each fund’s specific strategy. Not every good business is a good investment, and fund managers need to nail both.
Where We're Focused
Previously "too niche" opportunities are now viable at venture scale. Our focus areas include:
"Software+" models where AI makes hybrid approaches (software + physical data sources, software + human-in-the-loop) economically compelling, provided they trend toward software margins.
Proprietary data moats where multi-year data collection, proprietary data access, or exclusive partnerships create structural advantages that can't be easily replicated.
Regulated industries (finance, healthcare) where accuracy requirements, compliance frameworks, and workflow specificity demand more than general-purpose AI. Human-assisted verification remains necessary; specialized products that nail the workflow win.
Vertical-specific platforms positioned between legacy incumbents (technologically weak, often PE-owned, universally disliked) and horizontal tech giants (lack domain expertise). These startups have a wedge: legacy providers can't build modern tech, and platform companies don't understand the nuances. With a GTM-focused founding team and a thoughtful strategy, these can be winners.
If this resonates with what you’re building, we’d love to learn more. Email any of us directly or share your info here: bit.ly/laconiapitch
