Welcome to the 'Sherlocking' of AI
In the tech world, there’s a dreaded term for when a platform developer—like Apple or Google—builds a feature into its operating system that essentially copies and kills a popular third-party app. It’s called being “Sherlocked.” Today, the same phenomenon is playing out at warp speed in the artificial intelligence sector. Startups that have built clever products on top of OpenAI’s technology are waking up to find that OpenAI’s latest update, like the recent debut of GPT-4o, has integrated their core feature directly into the main platform. What was once a unique selling proposition becomes a free, native function overnight. This creates immense pressure. For months, founders could build a business by creating a sleek “wrapper” around OpenAI’s API—a
better user interface, a mobile app for voice conversations, or a tool for summarizing PDFs. Now, with OpenAI releasing its own desktop app, a more capable free tier, and advanced voice and vision features, the ground beneath these startups is shifting dramatically.
The Incredible Shrinking Value Proposition
The most immediate victims of this trend are the companies whose products are, in essence, just a thin layer on top of a large language model. Consider the ecosystem of startups that offered real-time, conversational voice AI. When OpenAI demonstrated GPT-4o having a fluid, human-like conversation—detecting emotion, understanding visual cues, and responding instantly—it leapfrogged many of these players. Their value wasn't the underlying AI, which they were renting from OpenAI, but the interface. When the landlord decides to build a better front door, the interior decorator’s job gets much harder. The same goes for countless tools designed for specific tasks like generating presentation slides or analyzing documents. As the base models become more multi-modal and capable, they absorb these use cases one by one. A startup that simply connected a user prompt to an API call is no longer a viable business; it’s just a feature waiting to be assimilated.
The Hunt for a Defensible Moat
So, how do you survive? The answer, according to venture capitalists and seasoned founders, is to build a “moat”—a defensible advantage that the platform owner can’t easily replicate. In the AI world, this moat is no longer the technology itself. Instead, it’s found in a few key areas. The first is proprietary data. A startup that has unique, high-quality data to train or fine-tune a model for a specific industry (like legal, medical, or financial services) has an asset OpenAI can’t just code into existence. The second is deep workflow integration. Instead of being a standalone app, the most resilient AI companies are embedding their tools directly into the complex software that businesses already rely on, like Salesforce, SAP, or specialized medical record systems. By solving a niche, enterprise-level problem and becoming part of a company’s critical infrastructure, they make themselves much harder to replace. The final piece is a truly unique user experience that creates a loyal community, something that a massive, general-purpose tool like ChatGPT struggles to foster.
A New Bar for Investment
This strategic pivot isn't just a founder’s headache; it’s fundamentally changing how venture capitalists evaluate AI startups. The era of funding a cool demo built on a GPT-4 API is over. Investors are now laser-focused on defensibility. Their due diligence questions have shifted from “What can your product do?” to “What happens when OpenAI does this for free?” They are looking for companies that aren’t just riding the wave but have a plan to navigate the increasingly turbulent waters. Founders seeking funding now need to articulate a clear strategy for building a durable business, not just a temporary feature. They must demonstrate a deep understanding of their target customer, a path to acquiring proprietary data or network effects, and a convincing reason why they will still be relevant after the next big keynote presentation. The bar has been raised, and only the most strategically sound startups will clear it.
