From: redpointai
Defensibility at the application layer in AI refers to the strategies and characteristics that allow a company to maintain a competitive advantage and resist disruption. This contrasts with earlier assumptions about defensibility in AI, which often focused on proprietary data or models [00:41:57].
Key Elements of Defensibility
1. Network Effects
Network effects are considered a paramount factor for long-term defensibility at the application layer [00:39:14]. Companies that prioritize building a strong multiplayer experience rather than just a single-player one are more likely to achieve this [00:39:14].
- Example: Chai Research Chai Research, a competitor to Character AI, outlasted its rival by focusing on network effects [00:40:02]. While Chai Research doesn’t possess its own proprietary models, it established a network of people submitting models to be run [00:40:18]. This marketplace approach, acting as a “choke point” or “protocol layer” between users and model providers, offers long-term robustness despite lacking proprietary IP [00:40:35].
2. Brand and Category Association
The ability to quickly become synonymous with an entire product category provides significant defensibility [00:41:20]. This allows companies to secure customer access and can even lead to increased Average Contract Values (ACVs), defying predictions of price compression [00:41:41].
3. Product Velocity and Breadth
The speed at which a company can develop and expand its product surface area is crucial [00:41:14]. Rapid iteration is essential, especially as new models are released every few months, creating existential events for companies that fail to adapt quickly [00:42:17].
4. Utility-Based Monetization
Applications can charge for the utility they provide rather than merely the cost of the underlying infrastructure [00:47:17]. This allows for higher margins and a more favorable business model compared to infrastructure providers, whose revenue is often reduced to “cost plus” pricing [00:47:21].
Contrasting Old and New Views of Defensibility
Historically, defensibility in AI applications was mistakenly associated with owning totally unique data sets or training proprietary models [00:41:50]. However, this has proven to be a “head fake,” as the real defensibility lies in the “thousand small things” that contribute to a delightful user experience, strong design, and rapid development speed [00:42:06]. This approach is reminiscent of how defensibility is understood in traditional application SaaS companies [00:42:29].
Defensibility in Enterprise Context
While network effects are critical, they can be more challenging to achieve in an enterprise context [00:41:06]. However, the benefits of velocity, product breadth, and brand hold true even in B2B settings [00:41:14].
Areas Where Defensibility is Emerging
The “first wave” of AI applications often focused on cost-cutting measures, such as automating tasks previously outsourced to Business Process Outsourcing (BPO) firms or managing customer support [00:32:04]. While these areas found early traction, they may face fierce price competition due to their perceived lower value [00:32:44].
The “next wave” of applications, focusing on top-line revenue generation rather than just cost savings, is expected to be more defensible [00:32:06]. Examples include:
- Outbound Sales: AI assisting with generating new revenue opportunities [00:36:05].
- Voice AI Applications: Such as scheduling intake companies, where even partial effectiveness (e.g., 75%) can significantly increase revenue for businesses that previously missed a high percentage of calls [00:33:17].
- AI education: Personalised teaching applications that may not be immediately obvious but have significant market opportunities [00:36:18].
- Finance: AI applications in finance.
These areas allow for a “clean ROI story” where customers are willing to pay a premium for solutions that directly contribute to increased revenue [00:33:06].