AI Bypassing SaaS: The Answer Isn't at Conferences

Let me share something. In Q1 earlier this year, IGV plummeted over 20%. The market blamed AI, saying large models and agents could directly generate code and execute tasks, bypassing SaaS and replacing seats. This logic sounds satisfying, but it feels more like an emotional purge. Recently, when helping clients break down their AI implementation checklists, the real bottleneck was whether enterprises had decomposed business actions into components callable by models.

Hundreds of CXOs attended the Digital Value Annual Conference. Titles discussed physical distillation and generalization, but the core contradiction was that individual case experience is valuable, yet scalable paths are scarce. CFOs and CIOs from Lingyi, iFlytek, CIMC, and Youngor talked about the last mile, while heavy-asset scenarios like Sany focused on organization, ROI, reliability, and talent. AI won't directly replace SaaS; it will first replace pseudo-digitalization that merely digitizes forms without ensuring data quality and process flow.

From my perspective, improved model capabilities don't automatically translate into enterprise capabilities. There's still a gap involving budgets, responsibility, system boundaries, and the actual operations of frontline employees. So I disagree with the linear narrative that "agents bypass software." A more likely path is SaaS transforming from interfaces clicked by humans to backends called by agents. The layer that keeps data, permissions, processes, and billing within the system becomes more valuable. The conference left answers undefined, which is honestly refreshing.