Last month, a moment highlighted a key issue for AI agent startups. When deploying Agentforce to follow up with over 1,000 previously neglected leads, the team found they could simply copy prompts from another AI tool and achieve excellent results immediately. This raises serious questions about the defensibility of AI agents. With 20+ AI agents in their toolkit from various vendors, the author notes that the differences between these agents are minimal. While some agents may have better user interfaces or specialized features, the core AI performance relies heavily on similar underlying models and training methods. The article breaks down what creates genuine competitive advantages in the AI agent space. Proprietary training data and custom models offer limited value; most agents leverage the same large language models. Instead, factors like network effects, deep integrations with platforms like Salesforce, and specialized infrastructure become more significant. The author emphasizes that building vertical solutions—targeting specific markets—can yield better results than general-purpose agents. Speed of innovation and creating lock-in through customer data also play crucial roles in maintaining a competitive edge. For buyers, this landscape offers some advantages. With the ability to easily replicate functionality across different platforms, switching costs are lower than assumed. The author suggests conducting vendor bake-offs to compare performance effectively. Negotiating hard is essential, as many companies overstate their AI's uniqueness. While generic agents face commoditization and feature convergence, highly specialized agents with unique workflows and datasets remain harder to replicate, highlighting a divide in the market.