The SaaS playbook has imploded: 10x your product or die

AI is changing how software gets built, and it is forcing a more fundamental question about the SaaS playbook. When building products becomes 10x faster, better, cheaper what looks valuable and defensible starts to shift. This is not a slow drift. The SaaS playbook as we knew it is breaking in front of us. The underlying economics and competitive dynamics have flipped, and both incumbents and startups are being re-rated through that lens.

We are entering a new era, AI or Die. Either make AI a source of customer value and operating leverage in your business, fast, or it will become the advantage your competitors use to dismantle yours. 

^{[Cheaper & Faster]}

My own 25 year experience: Building software is heading toward near-zero cost and lead time

I remember, when I launched La Nevera Roja - leading food delivery marketplaces in the 2010s in Spain - I asked a founder of a well-known wedding platform for advice. He told me that 10 years earlier, when they built what was essentially a simple wedding website, they outsourced it to a top-tier consulting firm for around €1.5m, and it took years to deliver.

When we built our site in 2011, a comparable product took roughly six months. A team with a deeper engineering bench could plausibly have done it in under three months, which still felt fast at the time and cost us roughly €150k, an order of magnitude in 10 years.

With the launch of Samaipata Fund I in 2016, we started seeing the next compression across the startups we analysed and backed. Modern frameworks allowed teams to ship meaningful products in weeks, and many early iterations moved from hundreds of thousands to the low thousands in cost. Another 10x compression in c.5 years

By 2026, or 10 years later, coding agents compress the loop again. Iteration has moved from weeks to minutes. The bottleneck is no longer building — it's upstream and downstream: figuring out what to build, and then getting it shipped and adopted by customers.

The economics back it up: fewer humans needed, and inference costs collapsing ~10x per year (a16z). Labor and compute are falling at the same time, the all-in cost of building software is heading toward zero.

^[Better]

When building is nearly free, you don't get less software. You get 10x better software.

When the cost of producing software collapses, the canvas becomes infinite. Teams don't just ship the same thing faster, they ship things they would never have budgeted for: deeper features, more polish, the extra 100 miles.

Anthropic built Claude Cowork in under two weeks with Claude Code. The point isn't the demo, it's what it implies: small teams can now deliver what only large, well-funded teams could before.

The bar moves up by 10x. And the data confirms it: app launches have accelerated to levels we've never seen.

What this means for traditional SaaS companies

Systems of record won’t be “vibe coded” from scratch. AI will still start eating SaaS from the edges, as new entrants ship workflow layers faster and at lower cost, often with better UX and iteration speed. The result is predictable. Pricing power weakens and growth gets harder to defend.

This is where margin pressure starts to bite. When software can be built dramatically faster, the supply of “good enough” products expands quickly and pricing power of incumbents weakens. The app-launch charts are a tangible early signal of that shift, with new releases accelerating sharply as these new development workflows become mainstream.

For traditional SaaS companies, the pressure comes from three fronts.

1. AI-native entrants that ship and iterate at a pace and cost base most incumbents struggle to match.
2. Fast-moving incumbents adopting AI to retrofit quickly and start competing on speed and perceived innovation both in their verticals and adjacent spaces
3. Customers revisiting build versus buy, as workflows that were previously uneconomic to build in-house become feasible with small teams plus agents.

The winners will be the companies that combine fast adaptation with real defensibility.

The critical step is translating AI-driven productivity into customer-visible value and aiming for a 10x step-change through better outcomes, better economics, or both. To do that consistently, it has to permeate internally as well, lifting output per employee and compressing execution cycles across the business. When that translation is slow, the sequence is familiar. Margins compress, retention weakens, growth slows, and access to capital tightens.

Defensibility matters because it is what protects pricing power as credible alternatives multiply. The strongest moats typically come from real switching costs, network effects and brand. A meaningful service layer can also become a moat, even if it used to be seen as a scaling drawback, because it is harder to replicate quickly than pure software. None of these moats are permanent, though. If you stop investing and evolving, they start to wear down, and standing still turns even a strong position into a shrinking one.

Make the transition real and make it fast. Translate AI into customer-visible value and operating leverage, and ship at the new pace before the market forces the change on you. 

For European founders, this is a great opportunity, as speed and product execution matter more than access to capital, and small teams can compete globally from day one. At Samaipata, we are leaning into that shift, backing AI-native companies already built on the new stack and supporting the rest of our portfolio as they accelerate their AI transformation.

‍

¹ Phrase borrowed from "AI or Die," essay by Ravi Gupta (Sequoia Capital), 2025.

‍