AI Startups Must Focus on Practical Implementation as Hype Alone is No Longer Enough

[baoshi.rao]

This year, it's rare to see investors offering help in times of need—most prefer to add icing on the cake. In the AI field, securing funding has become increasingly difficult, and the entrepreneurial journey is growing more challenging.

However, new trends suggest that the barriers to AI entrepreneurship are actually lowering. Solo entrepreneurs are flourishing in the wave of large models, appearing in every imaginable scenario—AI plugins, AI cameras, and more. More individuals are leveraging the convenience of large models to launch small-scale startups.

What’s evident is that AI entrepreneurship is shifting toward strong, practical, and implementation-driven applications. This aligns with observed trends: multiple AI entrepreneurs have emphasized that the prerequisite for starting a business is practical implementation.

For Chen Zheqian, founder of Yizhi Intelligence, who ventured into human-computer interaction as early as 2017, practicality is one of the key reasons their products are favored by many brands.

When discussing brand trends, Chen also noted that brands are increasingly focused on results, with ROI as a strong driver. Accordingly, their AI outbound calls and newly launched digital humans have adapted. More AI applications are emerging in marketing scenarios.

Jian Shi: The concept of the metaverse was hugely hyped last year. Is AI this year just another hype like that?

Chen Zheqian: Actually, AI didn't suddenly emerge today. On the contrary, it has already gone through several waves of development. From my perspective, I've experienced three waves of AI boom.

In 2012, facial recognition technology became a hot topic, and people realized that artificial intelligence was worth researching.

Then in 2016, the emergence of AlphaGo revealed even greater potential for AI. Human-computer interaction became a trend during that period, leading to the rapid emergence of various products like smart speakers. However, the technology wasn't mature enough at the time and didn't see widespread adoption.

Our company, Yizhi Intelligence, was boosted by this wave because AlphaGo can be considered the true starting point that sparked the AI boom in the entrepreneurial field.

So the large models that emerged at the beginning of this year represent the third wave in my view. Of course, if we look at the entire AI industry, this would be considered the fifth wave of boom.

Jian Shi: From 2016 to now, what do you think is the most critical factor in AI entrepreneurship?

Chen Zheqian: My biggest takeaway is that AI indeed has certain technical barriers—it's not something just anyone can do. If a high school student says they want to research AI, it’s highly unlikely.

However, AI isn’t so lofty that it’s unattainable. Some large companies are eager to portray AI as revolutionary, promoting the idea that it will transform human life. But at least based on their financial reports, the revenue clearly doesn’t match the scale of funding or the burn rate.

Big companies still have the financial resources to sustain AI development, but for entrepreneurs focused solely on AI, this isn’t a self-sustaining or commercially viable model that can create real societal value.

I’ve always believed that AI entrepreneurship must be grounded in practicality. I often tell my colleagues: we are fundamentally a business. We just happen to use AI technology to create a product. The key is to sell that product and ensure the profits are sufficient to support ongoing R&D and sales.

In recent years, AI has periodically sparked waves of enthusiasm. With each wave, some investors become willing to invest, prompting technically skilled individuals to start businesses even before their products are ready, often playing with concepts first.

However, this year's trends show that investors are far less likely to fund AI concepts. Everyone is being cautious, focusing more on the practical applications of AI. They won't provide help in times of need but will prefer to add to existing success. For entrepreneurs seeking funding, demonstrating tangible results has become crucial.

Jian Shi: So, there's no need to be pessimistic about AI. The most important thing is to have something substantial in hand.

Chen Zheqian: Exactly. I think a key point is that there's no need for some to dismiss AI or turn it into a grand banner. AI is, after all, just a technology.

If you hype a technology to great heights but fail to generate revenue, even the most advanced technology holds little value. Additionally, technology has its limitations. For example, while large models excel at image generation, they are slow in real human-machine conversations. If used for dialogue, they generate responses one by one, making you wait two or three seconds.

In phone call scenarios, people expect AI to respond within 500 milliseconds, not wait two seconds for an answer.

Technology will undoubtedly grow stronger in the future. But as AI entrepreneurs, we need to find Product-Market Fit (PMF) and truly commercializable scenarios within the current technological landscape.

Interviewer: How did you come up with the idea of AI outbound calling?

Chen Zheqian: From the start, my entrepreneurial direction was to develop intelligent dialogue systems. A dialogue system is somewhat like the brain—it needs a medium to realize its value. Some people integrate dialogue systems into smart speakers, TVs, or cars.

Back in 2017, looking at the overall level of the intelligent dialogue industry, smart speakers, autonomous driving, and AI customer service were still largely gimmicks. The robots at that time weren’t advanced enough to convincingly mimic real humans.

As I mentioned earlier, entrepreneurship requires considering feasibility. That’s why I chose AI outbound calling—because the phone itself is an excellent medium, and the brief duration of outbound calls (just a few minutes) allows the robot to convincingly imitate a real person during that time.

Further research into outbound calling scenarios revealed that consumer brands particularly needed this AI functionality. As a result, we’ve created value for consumer brand clients, achieved rapid revenue growth, and now hold a strong market share.

Jianshi: Why do you value human-computer interaction AI more?

Chen Zheqian: Human-computer interaction can bring businesses closer to users, thereby increasing users' recognition of the brand.

Technological iteration itself is a process of closing the distance. In the past, communication evolved from beacon fires and urgent horse messengers to watching movies from five to ten meters away, then televisions at two to three meters, and now smartphones within half a meter. With Apple's Vision Pro, the distance is reduced to just a few centimeters when wearing the glasses.

Hardware interaction devices are getting closer to human eyes. Similarly, marketing methods have evolved. In the past, media like newspapers only allowed one-way interaction. Later, in the era of self-media, users could comment and interact with brands via text. Now, for example, fans can watch live streams of celebrities and potentially get replies through bullet comments. AI outbound calls create a stronger sense of interaction with brands. This is why private domains are so popular—they bring users closer to the brand.

We have always used technological means to bring enterprises closer to users, thereby enhancing users' commercial value, or in other words, increasing their LTV (Lifetime Value).

Jianshi: During your outbound call operations, have you noticed any changes in brands' requirements for your AI capabilities?

Chen Zheqian: Definitely. What I've clearly observed this year is that many brands failed to meet their KPIs during events like 618.

Brands are increasingly focused on seeing direct conversions, with ROI as the strong driving factor, rather than planting seeds and waiting for them to grow slowly. Previously, we would do some phone-based product seeding, but now brands want to see faster conversions—for example, being able to see immediately after a call whether a user has made a purchase.

So now, brands aren't just asking how much AI can help reduce costs; they're more concerned with how much it can help increase revenue. This year, we launched a digital human tool in response to these changing brand needs.

Currently, we've noticed that brands are also very interested in large language models and digital human livestreaming. It's not just about saving costs on livestreaming for brands—it's also a revenue-generating tool.

The reason is that with the same budget, what used to only cover one live streaming room can now support multiple rooms with digital human technology. The difference in potential traffic between one room and ten rooms is significant.

Jianshi: Have any new job positions emerged within the company?

Chen Zheqian: Yes, in the digital human sector, we have a new role called Digital Human Live Streaming Operator. Their task is to manage digital human hosts and ensure smooth live streaming activities.

Unlike traditional live streaming operations where staff review sessions with human hosts, now they analyze performances with digital humans, researching ways to upgrade the digital hosts' dialogue and response mechanisms in live streams.

In fact, there are many new positions emerging in the market, such as AI Training Engineers. Additionally, I'd like to mention that besides AI entrepreneurs seeking funding, there's a growing number of grassroots AI entrepreneurs. These individuals leverage major companies' large models to start small-scale ventures in niche fields.

From the perspective of super individuals, entrepreneurship in the AI industry will become increasingly easier, not harder. This also means that in the future, more super individuals, or one-person companies, will emerge based on AI technology.

Jianshi: It seems you are a tech optimist.

Chen Zheqian: Because you have to believe that technology will ultimately bring about immense power. Just think back to over a decade ago when we didn’t have Baidu or Gaode Maps—it’s hard to imagine how people navigated without them. Similarly, ten years from now, we might not be able to imagine how we lived our lives today.

Technology will undoubtedly change many things, but we just can’t envision it yet.

Jianshi: From an entrepreneur’s perspective, what challenges might arise with the future trends of AI?

Chen Zheqian: I think there are three.

First, cash flow is the biggest challenge.
Cash flow depends heavily on initial investors. Companies that rely on investor funding rather than generating their own revenue will find it increasingly difficult to sustain operations in the future.

The second source of cash flow comes from customers. However, with many customers falling short of revenue expectations, cost-cutting measures are inevitable. This presents another challenge: how to sustain your business independently!

Don’t count on external support. In such tough times, businesses must fully mobilize their products, internal operations, services, R&D, and efficiency to ensure survival while maintaining high-quality offerings.

Second, competition will only grow fiercer.
The market is a battlefield. As the economy weakens, competition intensifies—everyone wants to survive, and no one wants to lose.

Startups already face immense challenges. To survive, they must rely on solid capabilities. However, some companies resort to unethical tactics, like price wars, when they can’t compete fairly.

This puts businesses in a dilemma: choosing between price and quality. The real challenge is: how do you stand out in such a competitive market and make customers remember you?

Third, pay attention to internal management. In the past, extensive management might have been acceptable, but under current circumstances, refined management is essential. Wages are one aspect—it's crucial to balance employees' psychological well-being. Additionally, spiritual motivation for employees is necessary, ensuring they feel they are gaining and growing within the company while receiving their salaries.