After seeing a thousand To B products, the youngest vice president of DingTalk started a business and created a non-To B Agent.

Text by | Deng Yongyi

Edited by | Su Jianxun

In 2023, Wang Ming had become the youngest vice president at DingTalk. He had contacted thousands of Chinese ToB SaaS software companies and enterprise customers. However, in his fifth year, he made a reverse decision - to start a business and develop a ToC AI-native application.

Wang Ming has had multiple career transitions. After graduation, he started a business in the "flash sale" retail sector. After the company was acquired, he went to 58 Group to be in charge of services like home delivery, always targeting the most ToC market. Then, during the COVID-19 pandemic in 2020, he joined DingTalk, responsible for the SaaS ecosystem, large models and AI ecosystem, industrial ecosystem, and AI innovation business.

Wang Ming said that during his nearly five years at DingTalk, having seen so many ToB products and models, he had a strong feeling that trends are more important than choices, and choices are more important than efforts.

The reason is that Chinese users generally lack the habit of paying for tools. For large customers, the migration cost of ToB business is extremely high, and the decision-making cycle is long. In the AI era, this is not the most suitable path for entrepreneurs.

However, when choosing the specific niche direction, Wang Ming set a premise for himself: to refuse to only develop tool-based products and to go global overseas from the very beginning.

Just one month after the establishment of Peak Intelligence, Wang Ming led his team to complete the first-phase construction of the new product, Moras AI, and also secured tens of millions of yuan in financing from Cloud Time Capital.

Moras is an AI Agent focusing on the content e-commerce scenario of influencers. In the first phase, it targets creators and individual merchants (prosumer) on TikTok.

The usage threshold has been minimized. Users only need to interact with Moras and express their preferences for the products to be promoted and the content. Moras will then automatically complete the processes of product selection, video script creation, business analysis, etc., and generate a promotional video of about 15 - 30 seconds.

Users only need to spend dozens of minutes a day learning know-how from Moras, interacting with it, and reviewing the content. Finally, they authorize Moras to search for products on TikTok, add them to the product showcase, publish videos, attach product links, write copy, select tags, choose channels, and then choose the appropriate time to publish the content.

"We don't guarantee that every video will go viral, but we can guarantee the overall ROI," Wang Ming said.

During the internal testing, Moras has partially verified this certainty. As long as users keep using it for more than a month, they can achieve stable sales. The current test data shows that some accounts have even achieved an ROI of 1:50 - for every $1 invested in computing power, it can bring $50 in GMV.

In terms of product form, Moras intends to shape the product to be more like a real person. From the virtual avatar used in the product to the overall product interaction form, it seems like a little assistant providing services to bloggers.

A small detail that reflects their aggressive product strategy is that the first version of Moras was only available as an app, with no web version.

Wang Ming said that he hopes users will think that they are really hiring an AI employee, and the usage threshold should be simple enough. "The web version is still more like a tool, and it's too difficult to use."

As for why he chose the TikTok market for the ToC product instead of others, it comes from his experience in operating platforms, two-sided transactions, and serving productivity users.

Overseas, the user base of TikTok is similar to that of platforms like Instagram and YouTube. However, the supply of commercialized content (such as promotional videos) and the bloggers' familiarity with the commercialization system are still in the early stages. The characteristics of users in developed countries determine that they don't prefer overly competitive business models.

"AI is not just a tool. The most important thing now is to directly deliver results." Wang Ming made a firm judgment to "Intelligent Emergence": "2025 is the first year of AI efficiency tools, and 2026 will be the first year of 'pay-for-performance' business platforms."

The logic behind this is that the emergence of AI can reshape industries, either through changes on the supply side or through changes in matching efficiency.

Wang Ming took an AI business travel platform he had incubated before as an example. Although it connected multiple suppliers through interfaces, making the enterprise ticket-booking scenario very smooth, it was essentially still a traffic business and did not change the supply cost or production relationship of air tickets and hotels.

But AI can do it. When incubating another AI legal service product, Wang Ming led his team to try a model where five AI Agents collaborated with real people. They provided legal counsel services that originally cost tens of thousands of yuan in the market at a price of only a few thousand yuan, and the response speed was faster.

In September 2024, after OpenAI's o1 brought the AI circle into the era of inference models, the evolution speed of model capabilities still exceeded everyone's expectations.

Becoming an Agent for content creation, product selection insights, and business data analysis is just the first step that Moras hopes to achieve. Wang Ming said that in the future, with the explosion of AI content, there will be a lack of trust. Personification will become an important anchor for trust in the business world. More and more business implementation actions will need tens of millions of "super individuals" with unique personas to drive them.

Moras hopes to become the Agent infrastructure platform for super individuals, help this group engage in lifestyle content e-commerce, and become the Shopify of the AI era.

For Wang Ming, what he needs to do now is clear: to quickly stand on the front line and feel the "tremors" of the models. "I didn't hesitate for a single second," he said.

The following is an interview between "Intelligent Emergence" and Wang Ming, the CEO of Peak Intelligence, edited and organized:

"The entrepreneurial opportunities for AI tools are over."

"Intelligent Emergence": Before starting your business, you had quite a rich experience. Could you elaborate on it?

Wang Ming: After graduation, I started my first business, creating something similar to today's flash sales. After two years, I sold the company. Then I started an internal venture at 58 Group, working on 58 Home Services. Two or three years later, I incubated 58 Enterprise Services. After another three or four years, I joined DingTalk in 2020.

At DingTalk, besides working on the SaaS ecosystem, I was also responsible for DingTalk's industrial ecosystem, enterprise-level services, large models, and AI application ecosystem.

Previously, most of my work was related to ToC and enterprise service platforms. Only during my time at DingTalk did I focus on ToB software. However, most of the time, I was involved in business models and trading platforms.

"Intelligent Emergence": You stayed at DingTalk for nearly five years. What was the moment that made you decide to start a business?

Wang Ming: Actually, the moment that inspired me was earlier. DeepSeek was a key turning point. But overseas, in the second half of 2024, when OpenAI's inference ability improved, entrepreneurship had already blossomed.

At the beginning of 2025, I was still at DingTalk. At that time, we all had a common understanding - in the future, there would definitely be many AI products that could deliver results. AI Native companies might first grow overseas, achieving a hundred-fold increase. After being able to deliver results (rather than just being tool-based products), they would then enter the domestic market.

"Intelligent Emergence": Moras is positioned as an "Agent that helps creators directly earn money." How do you understand what Moras does?

Wang Ming: Our idea is very simple. There are many users in the TikTok ecosystem. If they want to engage in commercial activities, they don't need to create content or select products. They only need to simply express their needs, then review - whether the content is okay, and then execute - publish the content when authorized.

In this process, what they need to do is authorize us with their accounts, and then review 5 - 10 pieces of content every day and click to publish.

It takes less than 10 minutes a day. But after a month, they can definitely achieve sales, which we have already verified through testing.

"Intelligent Emergence": What is the fan base of the users?

Wang Ming: In the initial stage, we focused on KOC influencers with a fan base ranging from a few thousand to tens of thousands.

"Intelligent Emergence": Before this interview, I searched for Moras. Its product form is quite aggressive, with only an app and no web version. Why?

Wang Ming: The web version still looks too much like a tool, and the threshold is relatively high. We don't want to create a pure tool. Our judgment is that in 2026, for startup companies, the opportunities for AI tools are over.

"Intelligent Emergence": Why do you have such a judgment?

Wang Ming: Why did most AI products lose money in the past few years? Of course, some AI products were actively pursuing a higher growth rate. But many AI products were still just tools. Even a powerful tool like Cursor, which can improve efficiency by thousands of times, although it had a fast growth rate, it did not bring certain commercial returns to creators.

Previously, no one was willing to pay for code completion tools. In the past, you were sold a hammer, while in the AI era, you are sold Thor's hammer, which can write hundreds of thousands of lines of code in a second.

But this still doesn't guarantee that you can make a fortune with this hammer. It's uncertain.

As a startup company, we need to find scenarios that are closer to money, more vertical, and can provide clear positive feedback on ROI. Even if you are working on a scenario with strong demand, if the penetration is still linear, it's not enough. We hope to achieve exponential penetration as soon as possible.

"Intelligent Emergence": What is your current business model?

Wang Ming: There are currently two models. The first is humans hiring AI - users pay the AI a basic salary plus a commission. I help you select products, create content, and achieve monetization. I will gradually teach you the logic of content e-commerce, acquire more context from creators, and in the future, become the creator's alter ego and persona to serve their fans well.

The second is AI hiring humans - the AI pays you a basic salary and can even negotiate the basic salary with you. Users help me review content and perform necessary actions.

"Intelligent Emergence": The second business model is very interesting. Why did you set it up like this?

Wang Ming: When collaborating with influencers, we found that there is a group of people who don't care much about the future ownership of IP and persona. They just want to do some work for the AI in the short term and earn money.

They said, "Can you pay me a basic salary first? I'll be your reviewer and executor."

This made me realize a disruptive possibility - when the cost of AI doing certain things is much higher than that of humans, we can hire them to help us build an account with a persona.

"Intelligent Emergence": Is the current maturity of the Agent sufficient to guarantee ROI?

Wang Ming: We can't guarantee that every product and every piece of content will be sold on every account. But on average, as long as users keep using it for more than a month, they can definitely achieve a lot of sales. We are currently conducting internal tests, and many KOCs have achieved results.

"Intelligent Emergence": If Moras covers enough processes, will it overlap too much with the platform? Won't TikTok launch similar tools itself?

Wang Ming: A platform can't be both the referee and the athlete.

You can't directly replace the work of influencers. Influencers will feel that the platform is taking over their jobs. The platform definitely hopes that third - parties like us will emerge and has provided us with a lot of support.

The Agent should operate as a platform and engage in two - sided business.

"Intelligent Emergence": You have worked in ToC (58 Group) and ToB (DingTalk) platform - type companies before. How did these experiences influence your choice of track and ToB/ToC?

Wang Ming: At this stage, if Chinese entrepreneurs want to become part of the global first - tier in AI applications, the only choice is to go global.

Let's analyze. As a Chinese team, if you choose to engage in ToB or ToG business right from the start, you can only rely on a white sales team to compete in the small and medium - sized enterprise market slowly. It will take one or two years to get orders from medium - sized enterprises.

But my judgment is that for AI application startup teams, the battle will end within a few years, and the giants of the next era will already show their dominance.

At this stage, if you are engaged in ToB business, you need to be like Palantir or Snowflake, getting orders from the world's largest companies in the United States. Each time you make a breakthrough, you can get tens of millions or even hundreds of millions of dollars in orders. Only in this way can you keep up with the first - tier in the AI application field in terms of capital.

This is not our strength. So we firmly chose to go overseas and didn't choose pure ToC. Instead, we chose creators in professional productivity scenarios, which are more like prosumers.

"Intelligent Emergence": Why did you choose TikTok instead of other ecosystems?

Wang Ming: As a startup company, we should find scenarios that are closer to money, more vertical, and can provide clear positive feedback to users in terms of making money.

TikTok e - commerce is such a scenario - you can help influencers sell products, and using the right content form to sell products is measurable and provides relatively fast positive feedback.

"Intelligent Emergence": There are quite a few competitors in the TikTok ecosystem doing AI applications, each cutting in from different angles. For example, Euka AI specializes in connecting North American influencers, Pippit AI focuses on video creation, and there are also teams like NemoVideo in China that start with video creation.

For prosumers, you are involved in content creation, product selection, business analysis, and direct conversion. Isn't it taking too big a step?

Wang Ming: At this stage, we haven't spent too much energy on the content creation process.

To be honest, what we initially wanted to create was the Manus in the e - commerce field - Agentic AI + infinite canvas.

Although AI coding is very advanced today, a truly good creative tool must be an Agent, and the interaction method should be an infinite canvas. There are still many points to polish, and the workload is huge. For a startup company, it's extremely challenging.

Moreover, after developing such a tool, many content creation tools can only be used by very professional creators and haven't broken into the mainstream. Or rather, they haven't been widely adopted by the majority of content creators because the threshold is still too high.

In 2024, we went on a team - building trip to Yunnan and saw a group of college students sketching. We made a bet and asked them if they had used AI. It turned out that none of them had. They weren't very clear about what Midjourney was, and they had only slightly heard of Doubao.

Today's AI users are lazier than we thought, and the penetration rate of AI is still very low.

"Intelligent Emergence": What is the approximate level of fineness of the content you can generate now?

Wang Ming: We can handle content with a few shots within about 30 seconds. In the future, we will gradually increase the number of shots and characters.

"Intelligent Emergence": To use AI - generated content to drive conversion and make this closed - loop work, what do you think is the most important thing?

Wang Ming: In 2025, many people want to create an AI version of Douyin. What does creating an AI Douyin mean? It means that the content creation threshold should be as low as that of Sora.

I thought about this a long time ago, when Sora 2 hadn't come out yet.

Sora 2 got it right in terms of content creation - the content creation threshold is extremely low. But Sora 2 was obviously developed by a team that doesn't understand business. An AI Douyin shouldn't just be a tool.