Founders must honestly assess if their product still creates a "jaw-dropping" reaction, similar to early experiences with powerful AI. If it doesn't, the product is losing its competitive edge and is vulnerable to disruption, regardless of existing customer contracts.
The current market leaves no room for mediocrity. SaaS companies are either at the forefront of AI, delivering jaw-dropping value and capturing new budget, or they are being displaced. Hiding behind long-term contracts is a temporary solution, as there is no longer a middle ground.
Data on 'vibe-coding' platforms shows that rebuilding a full SaaS app is an advanced, uncommon use case. Most users start with lower-risk, higher-ROI activities like rapid prototyping for engineering, building internal GTM tools, and automating personalized content creation.
Frustration with a mediocre, AI-lacking vendor drove the decision to build a custom replacement, even when a commercial option existed. This signals a major vulnerability for incumbent SaaS players who fail to innovate with AI, as customers may choose to build rather than renew.
With executive time valued at $1,000-$2,000 per hour, building a custom app that could be bought for $10,000 makes no financial sense. The justification to build must be a critical, strategic need for something unavailable on the market, not a desire to save on subscription fees.
The current focus in the AI-assisted coding space is on building apps. However, as more companies create custom tools, the critical, unsolved problem becomes who will maintain, update, and secure these apps over the next five years, creating a significant operational burden.
Advocates for buying most AI agents off the shelf to leverage existing solutions. Building should be reserved for the small fraction where no suitable tool exists, where you can replace a mediocre incumbent, or where proprietary data is a key advantage.
Desktop-based AI agents like Claude Co-Work, which can see your screen and local files, are a game-changer. They enable non-engineers to tackle complex projects like building production apps with single sign-on by providing real-time assistance and debugging.
To de-risk a complex technical build like implementing single sign-on (SSO), the speaker gave herself a one-day limit. If the core technical hurdle wasn't solved within that day, the plan was to abandon the custom build and revert to the paid, off-the-shelf tool.