Competing with giants, the end of SaaS & other listener questions

REWORK host Kimberly Rhodes asks 37signals co-founders Jason Fried and David Heinemeier Hansson whether agent-built custom software will end SaaS. They see the near-term threat as uneven: expensive, disliked enterprise deployments are exposed, while most low- and mid-market customers still pay vendors to own the maintenance, operations, and product decisions.

Key Points Covered

• Compete for a viable niche, not to defeat a giant: Fried tells small builders to focus on their own costs, customers, and reason to exist. The immediate test is whether a few customers will pay, not whether the product can displace Google or Apple [00:01:02]-[00:03:04].
• Use smallness as differentiation: Hansson suggests asking whether you would buy your own product over the established alternative, then leaning into advantages a large company cannot copy, such as direct access and a narrower, more distinctive product [00:03:04]-[00:05:12].
• Replacing SaaS with agents is still pioneer behavior: Hansson says most examples he sees come from people who already want to experiment with agents. He sees greater near-term exposure in expensive, disliked enterprise systems than in inexpensive tools for small companies [00:05:12]-[00:07:55].
• Buying software transfers operational responsibility: The barrier is not only building a replacement. Customers must run it, maintain it, manage its data and backups, and preserve ownership when employees leave. Hansson points to limited uptake for 37signals' self-hosted ONCE products as evidence from their own business, not a market-wide measurement [00:07:55]-[00:09:42].
• AI adds another competitive option rather than creating churn from nothing: Fried notes that customers already switch because of competitors or weak products. Agent-built tools add one more alternative, so vendors should assess their actual price point and customer segment instead of treating SaaS as one market [00:09:42]-[00:10:44].
• Build-versus-buy economics still matter: Fried contrasts a hypothetical million-dollar replacement with recreating a product that costs $600 per year. Even with agents, employee time, maintenance, breakage, and opportunity cost can exceed the subscription while leaving no vendor accountable for support [00:10:44]-[00:12:47].
• Implementation is only part of a product: Fried argues that strong software still needs a cohesive concept, boundaries, and an understanding of what customers are trying to accomplish. Faster code generation helps established vendors too; it does not remove the product work [00:12:47]-[00:14:30].
• The larger shift would be agents choosing the solution: Hansson distinguishes implementing a customer's request from diagnosing the problem and deciding what software should exist. He has not yet seen autonomous product direction work well; his strongest examples still require a person with taste, vision, and domain knowledge to steer agents [00:14:30]-[00:16:38].
• Cheaper implementation creates new human competitors: Hansson sees a nearer-term threat in people who already have product taste and ideas but previously lacked programming ability or funding. Agents let more of those builders turn their ideas into working software [00:16:38]-[00:18:29].
• Personal presence works better when it is not manufactured: In the final question, Fried and Hansson reject formulaic “personal brand” tactics in favor of talking plainly about work and interests. Hansson also acknowledges the tradeoff: ordinary corporate social accounts often attract little attention, forcing founders to mix personal and company communication [00:19:29]-[00:27:50].

Full video: https://www.youtube.com/watch?v=wpdi1bp-EbU(opens in a new tab)