People & Blogs

AI做软件这么强了，SaaS何去何从？

by 晓辉博士

📚 Main topics

AI is making software easier to buildOpen-source tools and AI can now recreate or add software features that once required substantial development work. This raises the question of whether customers will continue paying for SaaS. 0:00
Why a manufacturer is building its own toolsOne manufacturing company has largely stopped buying external SaaS, using AI to build customized software and keep company knowledge and communications under its own control. 1:00
An independent developer’s changing expensesLevels.io says he has canceled his business SaaS subscriptions, with remaining costs focused on AI inference, hosting, storage, and backups. 3:06
Where SaaS can still deliver valueCustomers may still pay for services that reliably deliver outcomes, handle complex backend work, or meet compliance requirements. 4:07
The shift from software features to agentsSaaS companies may increasingly offer agents that perform tasks reliably, rather than simply selling access to software features. 5:09

✨ Key takeaways

Customization and control matterBuilding software in-house can help companies tailor tools to their needs and retain sensitive information, while AI makes development faster. 1:31
Reliability is a differentiatorBusinesses may prefer a provider that takes responsibility for delivering a dependable result over assembling and maintaining their own AI tools. 4:07
The choice is employee, in-house AI, or third-party agentCompanies will compare the cost and quality of each option, including whether an outside provider offers a safety net if something goes wrong. 6:11
Token economics will shape decisionsCustomers will need to understand the cost of compute and the value an agent’s work creates. 5:40

🧠 Lessons learned

Software features alone may not be enoughIf a function is easy to recreate, SaaS providers need to offer more than the feature itself, such as dependable service or hard-to-rebuild infrastructure. 4:07
Human strengths depend on the taskAI may be more effective for clear, standardized work, while people remain valuable where aesthetics, judgment, communication, or complex decisions matter. 6:42

🏁 Conclusion/next steps

SaaS is likely to be restructured, not disappearThe future may favor agent-based services, with companies weighing whether to buy them, build their own AI solutions, or assign work to employees. 5:40

Transcript excerpt

0:00 Hello, today we'll discuss the impact of powerful AI on SaaS and the future path for software companies. Over the past few days, there has been a lot of discussion on X and in local communities about reverse engineering. A recent article mentioned someone who reverse-engineered the entire Adobe software suite, essentially using AI to rewrite this incredibly complex software and open-sourcing it for everyone. There’s also an open-source framework called Rear that can help you

0:30 reverse-engineer many applications. I even tried it today to help me write some simple functions. For instance, features like beautification or adding subtitles; I had already covered almost everything while polishing my skills, except for the beautification part. But today I discovered that, yes, it can be done; in fact, many open-source frameworks can help you execute these tasks now. This leads to the question of whether people will still be willing to pay for these SaaS products in the future, which I think is a big question mark. Actually, a couple of days ago, I

🔒 The full, searchable transcript is available with Pro.

Suggestions