Building a pure technology moat has become challenging since the emergence of LLMs. Due to the lower barriers of entry for introducing new products to the market and the continuous fear of becoming outdated overnight, investors and startups are trying to find a path to sustainable competitive advantage. However, this new landscape also presents an opportunity to establish a different kind of moat. A moat based on a much wider product offering solving multiple pains for your customers and fully automating large workflows from start to finish.
The AI explosion, whose blast radius has kept growing since the public launch of GPT3.5/Chat GPT, has been mind-blowing. In addition to the discussions around efficiencies and risks, startups in the space and VCs found themselves dealing relentlessly with the question of whether building a technology moat is still possible. Companies are struggling with the realities of creating a defendable product that has substantial barriers to entry for new competitors or incumbents. Just as in the past, this will continue to be a necessary component for a startup to be able to grow and become a Centaur or Unicorn.
The real revolution isn’t just Chat GPT. The real revolution includes open-source models becoming available for commercial use – for free. Additionally, solutions such as LoRA are allowing anyone to retrain open-source models on specific datasets quickly and economically.
The reality is that while OpenAI kicked off the era of “Democratization of AI”, the open-source community kicked off the era of “Democratization of Software”.
What this means for startups and VCs is that now, instead of defining narrow, “single-feature” products that solve niche pains that have remained unmet by competitors, startups can listen to their customers on a much broader scale and deliver wide products that solve multiple pains that seemed unrelated only a year ago. When combined with integrations that fully automate customers’ workflows, startups are able to truly achieve sustainable competitive advantage.
Simply put, startups will need to connect the dots between problems, find solutions that no one else has considered, then find additional dots to connect.
Put yourself in your customers’ shoes, when they are searching for a solution for a certain pain; they are finding dozens of solutions at a time. How do you go about understanding the differences and evaluating them? How can you make long-term decisions if you feel there might be more solutions available next month?
Customers would much rather have one “AI partner” that updates its offerings with the latest tech, rather than multiple small vendors.
Executing this strategy requires setting a broad vision and much shorter, targeted cycles across the organization both in product development and for company-wide synchronization. For instance, ML/AI teams should be part of weekly sprints. This will allow them to more efficiently add new AI features and make decisions regarding adding new LLMs or open-source models within the same time frames to either improve or enrich offerings.
By building a wide product instead of one focused on a single feature, startups can achieve this mythical moat since it simplifies product adoption, creates further barriers to entry (against both new entrants and market leaders), and safeguards against new open-source models that could be released and tear down a business overnight.
Let’s take a look at the AI transcription market (ASR) for example – several providers were in this market with similar price levels and relatively nuanced product differentiations. Suddenly, this seemingly sleepy market was rattled when OpenAI released Whisper, an open-source ASR, which showed immediate potential to disrupt the market but with some substantial gaps. The “incumbents” in the market, who faced the above dilemma, decided to each launch a new proprietary model and focus some of their messages on the problems of Whisper.
Meanwhile, others found ways to close these gaps and market a superior product with limited R&D efforts that is receiving incredible enterprise customers feedback and an entry point with happy customers for expanding into our full offering.
Returning to the original question – can one build a moat in the AI space? I believe that with the right product vision, agility, and execution, startups have the opportunity to build rich offerings and, in due time, compete head-to-head with market leaders. Many of the core principles needed to identify great startups are already inherent in the minds of VCs who understand what it takes to recognize opportunities and grow them accordingly. However, it’s critical to recognize that today’s castles look different than they did years ago. What you protect is no longer the crown jewels, but the whole kingdom.