New Characteristics of AI Entrepreneurship
-
There have been many comparisons between AI and mobile internet, but after actually participating in some AI projects, I found that the entrepreneurial patterns of AI and mobile internet are quite different.
First, the validation period for AI startups from 0 to 1 is longer. When developing mobile applications, we can simplify the product to a single feature, create an MVP in two to three months, and launch it for validation. Mobile applications are confined to specific scenarios. For example, if we develop a ride-hailing app, due to its product positioning and specific interaction logic, users will only use it for ride-hailing and certainly won't search for news within it.
However, AI products are different. When users encounter a chat interface with AI capabilities, they will input any content to test its boundaries. If the product fails to handle these inputs, users are likely to leave. To some extent, the concept of MVP is ineffective for AI products, or rather, the time required to build an MVP for AI products is extended.
Mobile products typically address users' deterministic needs—ride-hailing is for transportation, food delivery is for meals. In contrast, AI products face users' uncertain demands. When presented with an input box, users may discuss everything with the model, making expectation management a crucial issue. At the beginning of the year, people marveled at ChatGPT's capabilities because demos easily showcased the upper limits of the model's abilities. However, in real-world usage, retaining users depends on the lower bounds of the model's performance.
Secondly, once Product-Market Fit (PMF) is correctly identified, the growth trajectory of AI startups from 1 to 10 accelerates significantly. In previous platform-level entrepreneurial waves like PC and mobile internet, the proliferation of applications was constrained by the pace of hardware adoption—developing both terminals and applications simultaneously. In contrast, AI startups build applications on existing smartphones and PCs. As long as the product is remarkable, it can achieve viral adoption at lightning speed. Cases like ChatGPT and Midjourney, which reached hundreds of millions of users within just a few months, have already validated this pattern.
However, because every terminal position is already occupied, with no traffic dividends left, this demands that new AI products must be ten times better than their predecessors. Ideally, they should solve problems that were previously unsolvable to have a chance to stand out.
In the early days of PC and mobile internet entrepreneurship, the business infrastructure was immature - there were no ready-made payment systems, monetization methods, or traffic distribution channels. During the mobile internet's infancy, people had no idea how to make money on mobile devices; concepts like feed stream advertising were unimaginable. Today in AI entrepreneurship, these business infrastructures are readily available, but the core 'engine' - the AI models - remains immature.
Here's what will happen: some smart people will first use a 50-point AI product for practice, start commercialization immediately upon launch, explore growth methods, and thoroughly understand the product's essence through practical experience. When the foundational model reaches maturity (its '18th birthday'), they'll replace it with a new engine and take off directly. Unlike the internet era where companies purely focused on user acquisition without monetization models and burned cash recklessly, this scenario is unlikely to happen in AI entrepreneurship - AI applications inherently come with monetization models.
Third, major competition arrives earlier, and for startups, the first battle may be the decisive one. Back in 2010, few major companies prioritized mobile-first strategies. Even by 2012, some were still hesitant about the opportunities in mobile internet, which gave startups time to grow under the radar. By the time their strategies became clear, these startups had already grown too strong for the big players to suppress. Examples include ByteDance challenging Baidu, Pinduoduo and Meituan taking on Alibaba, and miHoYo competing with Tencent. These startups moved from the periphery to the core, overturning the established order and forcing big companies' management to adopt a VC-like mindset—believing in the power of non-consensus ideas and not underestimating seemingly small entry points.
Therefore, the consensus among major tech companies on AI came much earlier. Almost all large corporations regard AI as a new growth curve, intensifying competition with startups ahead of schedule. In AI entrepreneurship, one thing is certain: once the 0-to-1 validation is achieved, the next one to two years will become particularly critical. The growth from 1 to 10, the rollout of commercialization, and the development of organizational capabilities will all occur within a very short time window. Previously, completing this entire process might have taken four to five years, but now it will be compressed to two to three years. This demands not only the founder's initial momentum but also sustained acceleration.
Previously, in a discussion with Qu Kai from 42 Capital, we reached a consensus that higher talent density and a more fiercely competitive environment will accelerate the arrival of 'AI 2012'. Years from now, looking back, we will still witness a group of startups at the main table, challenging the industry giants.