Large Models+: What New Entry Points, Changes, and Possibilities Lie Ahead? | Yunqi Capital · Riding the AGI+ Wave

To be updated)

As natural language becomes the new medium for human-computer interaction, it offers not just new ways to unlock productivity, but opens the door to unleashing imagination. In our previous "Riding the AGI Wave" special, we explored key considerations for software companies integrating large models. Innovation in content and workflow streams, while dramatically boosting efficiency, may also give rise to entirely new business paradigms.

In the new practice of embracing large models, what new entry points, changes, and possibilities lie ahead?

(Surprise bonus at the end of the article — don't miss it!)

Feng Yao, Managing Director

We're now seeing two modes of thinking: one seeks optimization within existing frameworks, such as using AI to improve efficiency in sales, supply chain, delivery, and other individual links; the other completely discards existing structures to find entirely new architectures from the perspective of how technology can disrupt. We hope entrepreneurs can think in both modes, and look forward to fresh insights emerging from the collision between the two.

Xu Zhixing, Frontier Tech Investor

Scenarios that have already completed digital closed loops will gain more pronounced advantages in this AI evolution; industries with better data accumulation will see more tangible output effects in a short time.

Where are the "new entry points"? The production threshold for consumable content is being lowered. With the boost of AIGC, products like Midjourney and Stable Diffusion are reshaping the design industry. From "creative idea" to "finished product," efficiency rises and costs fall, users have more choices, and "creativity" may become a new commercial entry point. In the future, industry-grade solutions with more refined know-how may make "what you see is what you get" flexible manufacturing a reality.

We've also observed that although the large model ecosystem is being further developed, the greatest commercial value is more likely to be absorbed by the two ends of "infrastructure layer" and "traffic entry points." On the "traffic entry points" end, the internet has already solved the "information barrier," and large models have gone further in information search, summarization, and induction, solving the "skill barrier." After development thresholds are lowered, internal company model deployment can quickly generate business tools, so SaaS companies need to have a more thorough understanding of the full service process, find critical traffic entry point nodes, connect upstream and downstream products, and provide "systematic solutions" to deeply bind with customers.

What are the "new changes"? Business essence and industry characteristics determine application scenarios. In industries with existing demand but lacking appropriate technology, product forms are most likely to be rapidly and dramatically reshaped.

Take CRM as an example: using AI to record sales leads and generate searchable content can improve the accuracy of the records themselves, and describing them in natural language will greatly enhance user experience. In scenarios with existing data accumulation such as online customer service and B2B marketing, our early portfolio companies Sobot, Weiling Technology, and Bailing Intelligence have already quickly launched multiple AI applications.

For industries with inherently complex task models and high data irreplaceability, directly接入ing large models at this stage yields less impressive results, but large models can assist with data analysis and other tasks. We've also noticed that faced with highly specialized questions, ChatGPT is becoming increasingly "smart" — ChatGPT-4's scores on biology exams have far surpassed ChatGPT-3.5's.

Changes in business models will also bring adjustments to company talent structures. AI won't directly "replace" current jobs, but will more often take the form of a co-pilot, better assisting humans in completing work. Technology iterations will also spawn new "job types" — we've already seen "prompt engineer" become one of the most scarce talents overnight.

What other new possibilities are there?

Although technology development has appeared to advance at breakneck speed over the past few months, in practice, large model development and applications are still in early stages.

Returning to the actual circumstances of SaaS companies, multiple entrepreneurs have raised in conversations that change cannot happen overnight, but occurs gradually with warmth and tactile understanding. SaaS remains inherently a "slow business," with iteration built upon layered, progressive business understanding. This "warmth" also extends to offline spaces, serving as an anchor of constancy amid change. The source of all "new changes" must still return to the original intention of helping customers solve real problems.

At the event, multiple entrepreneurs shared their firsthand experiences in embracing new technology —

Tang Jinwu CEO, Langjing Technology

AGI as a tool will greatly improve company operational efficiency. But the essence of the retail ecosystem hasn't changed — no matter how AI technology advances, human connection in physical space is irreplaceable.

Hua Shao Chairman, Cool College (Kuxueyuan)

A company's core moat still lies in the value its product delivers to customers. Only by embracing the changes AGI brings and developing new products according to customer needs can competitiveness be improved.

Yu Wei Chairman, Ruiqi Technology

Technology has always been about solving real-world needs. Traditional industries already had a solid foundation in the mobile internet era; timely embracing new technology for transformation will better improve efficiency and service quality.

Li Jianghua Founder & CEO, Yundongyun

In the past, ToB products were basically menu-driven. The first intuitive change AI brings is the transformation of interaction modes — menus may be directly replaced, with users awakening the functions they need through conversation or voice.

Chen Zhigang Founder & CEO, Hectan Pharma

In the future, the alignment of multi-disciplinary knowledge will be an important application direction for AI, as well as a challenge and opportunity for startups. For example, in cross-disciplinary reasoning, AI can mobilize external models to generate more unexpected applications.

What other new possibilities await verification? The "Yunqi AGI+" series continues this summer — this time, we'll partner with renowned universities, frontier entrepreneur communities, and seasoned industry experts to enter the Greater Bay Area, dialogue with tech entrepreneurs in Guangdong, Hong Kong, and Macau, and jointly seek innovative sparks at the intersection of AGI and industry.

Event registration is now open — looking forward to meeting more entrepreneurs in the Bay Area!