Can "pay-for-results" save enterprise software?

Enterprise software in the AI era will ultimately move towards role - based, result - oriented, and value - closed - loop models.

In the circle of Chinese enterprise software, almost every founder working on AI is thinking about one question: Can my product be charged only based on the "results"?

Customers are no longer willing to pay for tools. They want clear returns such as "efficiency improvement", "revenue increase", and "cost reduction", and preferably, they can sign a performance - based agreement. Result - orientation seems to have become the threshold for survival in the next era.

However, the problem is, how exactly is the "result" defined? Can it be delivered? Will customers recognize it? This cannot be solved by simply saying "pay based on results".

Meanwhile, "pay - for - results" seems to be becoming the most controversial and anticipated new paradigm in the field of enterprise software.

At the recently held 2025 AI Cloud 100 China list release event, Cui Qiang, founder & CEO of Ciniu Club hosted a round - table discussion with Chen Hang, co - founder & CEO of Qunhe Technology; Jin Lijian, founder & CEO of Yingdao RPA; Sima Huapeng, founder, chairman & CEO of Silicon Intelligence; and Zhao Chong, founder & CEO of PixelBoom (AiPPT.com) to deeply explore "From software subscription fees to paying for GenAI results, how will AI reshape the revenue model?".

Cui Qiang raised the core question at the beginning: Is it destined that 99% of enterprise software cannot achieve the "pay - for - results" model? And what exactly does that possible 1% look like?

For AI enterprises, the question is not "whether they can charge", but "for whom to deliver, what to deliver, and how to collect money".

Chen Hang believes that the core is not the model, but whether customers truly recognize the value of your product. As long as customers approve, a suitable monetization method can always be found. He emphasized that the essence of AI implementation scenarios is the matching between efficiency and output. "Subscription" and "results" are actually just two means.

Jin Lijian proposed that the role of AI should no longer be just a simple tool, but should become an integral part that can participate in enterprise processes and even organizational decision - making. However, the premise is that you must clearly define the boundaries of the "result"; otherwise, it will be difficult to form a reasonable business closed - loop.

Zhao Chong provided another answer: Between ToC and ToB, he chose the entry point of the "creative expression" scenario and explored the generalized result delivery with a mixed - payment model. Even if it only saves 2 hours of PPT - making time, it is a perceptible result.

Sima Huapeng stands more firmly on the side of "pay - for - results". He compares AI to an "executive" who can bet on performance and emphasizes that only by deeply participating in the industry and sharing business KPIs can AI enterprises obtain truly high - value returns. He once said bluntly, "An AI that can't help customers earn one million yuan is not a good AI."

The final discussion among the five guests returned to the most fundamental questions: What is a "good result"? Who will judge it? And how should it be priced?

Facing these questions, the answers may not be unified yet, but a consensus is being formed: Enterprise software in the AI era will ultimately move towards role - based, result - oriented, and value - closed - loop models.

And "pay - for - results" may be the door to the new paradigm.

Table of Contents

1. Not all software can be "charged based on results".

2. Will pay - for - results move into the "deep water area"?

3. The underlying challenge of pay - for - results: Who will define the value?

4. Who decides what a "good result" is?

5. From individual cases to the mainstream: Is the future of pay - for - results here?

The following is the edited and organized dialogue content by Neuters:

Not all software can be "charged based on results"

Cui Qiang: The topic we are discussing today is whether AI will reshape the traditional software subscription revenue model. Yesterday, I talked to a group of entrepreneurs in Hangzhou about a similar question, and we reached a conclusion: Approximately 99% of enterprise software cannot achieve the "pay - for - results" delivery model. Only about 1% may truly be able to do so.

Some guests mentioned in the previous sharing that the advertising industry was actually an early representative of pay - for - results. So, the question is: Did AI give birth to the development of pay - for - results, or did pay - for - results already exist, and AI just accelerated its implementation?

Additionally, I also want to invite each guest to share: Have you tried to transform the subscription model into a pay - for - results model? If so, what was the effect?

Chen Hang: We are working in the field of spatial intelligence. Our most well - known product is Cooljia, which has now expanded to multiple industries such as home design, chain stores, e - commerce space design, industrial manufacturing, and intelligent agent training, and our service targets cover various types of customers.

The first element is user value: Whether customers are willing to use your product is the core. Only when "1" holds true, that is, when customers truly recognize you, can the subsequent business models be flexibly adapted. In the end, as long as users are willing to pay, there is always a way to charge.

After the emergence of AI, many companies, especially those involved in 3D training and video generation, face the problem of very high supply and model costs. If you want to achieve a high - quality result, you also have to pay a high cost for cloud services. In this case, charging based on "results" is reasonable, just like OpenAI itself.

So, it's still the same thing: "1" is more important, and the model comes second.

We serve many industries, and the charging method actually depends on several key factors. The premise is to assume that your marginal cost is not particularly high, or it can be regarded as a constant:

If the customer base is large, for example, targeting a large number of salespeople, the subscription model is more suitable;

If the number of customers is small, then it depends on the effect, and pay - for - results may be more suitable.

The core of AI's involvement is still whether it truly improves efficiency. It also brings an important distinction:

Does your product solve marketing problems or cost problems?

If it is just a cost center, it is difficult to promote long - term subscriptions from customers, and pay - for - results is more likely;

If your product is a marketing tool that can bring more output from salespeople and even help the enterprise expand its scale, then the subscription model is more likely to be established.

Another dimension is the user attribute: ToC products are more likely to implement pay - for - results, while enterprise software for ToB requires more complex persuasion and logical self - consistency.

So, in summary, the three major factors to consider in the pricing model are:

1. Is the number of service targets large or small?

2. Is it targeting a cost center or a marketing center?

3. Is it ToC or ToB?

Jin Lijian: Our product, Yingdao RPA, initially achieved automation through logical rules to replace some repetitive work. After the emergence of AI, its ability boundary has been broadened. It can not only execute rules but also handle some decision - making tasks.

From the customer's perspective, they always pay for "results". Therefore, during the sales process, we will deeply understand the actual scenarios of enterprises, learn about the daily work processes of employees, and help customers identify which links can be replaced by Yingdao. This is essentially a result - oriented sales logic.

Although AI now has stronger capabilities, our current charging model is still mainly subscription - based. The overall logic is: Sell our products around the results expected by customers to achieve charging.

Cui Qiang: Generalize this problem through the model.

Zhao Chong: We are working on an AI Office suite. Our core product is AiPPT.com, and we also have other products such as AI Table (AiBiao.cn) and AI Note - taking (AiHaoji.com).

The biggest difference between us and the previous generation of tool - type products is that we directly deliver results rather than just providing tools. Traditional PPT tools give the tool to users, and users still have to find pictures and create content by themselves. But we directly help users generate content and create PPTs - we deliver complete results.

In terms of the monetization model, we adopt mixed charging: The basic functions are subscription - based. You can generate thousands of PPTs within a year for 119 yuan.

However, if users have more professional needs, such as:

Access to specific knowledge bases (for example, in the government and enterprise version, access to mainstream value - based corpora and content review)

Enterprise - customized VI brand color schemes, Logos, fonts, and enterprise - customized templates

Advanced collaborative functions

These all require separate payment.

That is to say, we focus on "result delivery" and adopt a differentiated and mixed monetization model in different usage scenarios.

Sima Huapeng: We are Silicon Intelligence, working on digital human products. Compared with the other guests here, the unit price of our products may be the highest. The average price of our AI live - streaming products is between 60,000 and 100,000 yuan per set.

I am very optimistic about the pay - for - results model and have done a lot of practice. I believe it is the only way to truly break through to AGI.

Why can we sell our products at such a high price? Because many customers have created huge value with our tools.

Let me give you an example: This year, we provided AI capabilities to a live - streaming company with only a dozen employees. The company's AI live - streaming revenue is expected to reach 100 million yuan this year. We are also trying to obtain an appropriate proportion of the revenue through in - depth cooperation.

As we all know, Luo Yonghao recently had an AI live - streaming event, and it is said that he sold more than 50 million yuan worth of goods in one event. Coincidentally, he was also at the forum yesterday.

But my question is: If this is really a tool that can bring tens of millions of yuan in revenue, how much should it be sold for? How much subscription fee is reasonable? Obviously, participating in the revenue distribution is more reasonable.

Since last year, we have been continuously exploring the "pay - for - results" path, and it has been successfully implemented, especially in the live - streaming scenario. A foreign customer sold hundreds of millions of dollars in GMV using AI in a year, and we were able to lock in a proportion for revenue sharing. This is a direct manifestation of pay - for - results.

In addition, we have used AI to create multiple online celebrity IPs and directly use the traffic of these IPs to help customers generate revenue. Since April, I have spent two months using our AI capabilities to "transform" myself into a knowledge blogger with nearly ten million followers. Now, the quotation for a brand advertisement on this account is 150,000 - 200,000 yuan, and the advertising revenue is expected to reach tens of millions of yuan this year.

This is also a transformation from selling tool subscription fees to selling results, and I am a special customer in this case.

During the upsurge of AI large - models in 2023, I once put forward a view: "An AI that can't help customers earn one million yuan is not a good AI." If your AI product can only be sold for 19 dollars a month, the value it creates should not be much higher than this price.

We dare to sell some of our products for tens of thousands or even hundreds of thousands of dollars because we can clearly see the direct benefits that customers can obtain from using AI. Only an AI that can truly achieve pay - for - results can be called a "valuable partner".

We will continue to pursue this path alone.

Cui Qiang: Current AI products not only deliver tools but also need to provide consulting services, and even accompany customers closely and get directly involved in helping them. This has already deviated from the category of "tools" itself, hasn't it?

Sima Huapeng: The value created by pure AI tools will, with the intensification of competition, lead to lower and lower prices for customers, and may even become free and open - source.

Just like Wang Ning, the founder of Pop Mart, recently said, all "useful" things will eventually be involved in price wars because "usefulness" itself means high substitutability.

On the contrary, Pop Mart focuses on "uselessness". For example, Labubu, although it has limited functionality, brings huge emotional value and investment attributes, and you may be able to make several times the profit by reselling it.

So, returning to the essence of AI competition, we should avoid getting involved in price wars and instead truly create value for customers and obtain revenue through result - sharing.

For example, once autonomous driving technology matures: Will you operate a taxi company like Tesla, or will you sell the technology? The answer is obvious.

Another example is that if you develop an AI stock - trading software that can make money from stock trading, will you sell the software directly, or will you use it to trade stocks by yourself? You can get the answer with the most basic economic common sense.

Instead of boasting about how much efficiency your AI can improve, it's better to get directly involved, use it to help customers make money, and finally participate in the profit - sharing and deeply participate in industry transformation.

We have already implemented this in fields such as short - video, live - streaming, and emotional companionship and have established a large number of self - operated businesses. It is expected that in the future, self - operated revenue will account for two - thirds of our total revenue, and the proportion of traditional software sales revenue will gradually decrease.

Will pay - for - results move into the "deep water area"?

Cui Qiang: When I invited these four guests for this dialogue, I suddenly realized that their products are mostly at the tool level and hardly involve the management level. This may also represent the views of some people.

What does the widely - used concept of "pay - for - results" really mean? Pay - for - results may ultimately mean getting deeply involved in every link of the industrial chain. Once that happens, things will become complicated, and it will be difficult to maintain the rapid replication characteristic of tools.

I want to ask: Will the real pay - for - results in the future develop in the direction of going deeper into the industrial chain?

Zhao Chong: Pay - for - results has many levels. It doesn't necessarily mean getting deeply involved in the customer's business to help them make money to be considered result delivery.

For example, some of our customers have a high customer unit price, reaching hundreds of thousands of yuan, which is for result delivery in scenarios such as improving government office efficiency. Obviously, it's unrealistic for us to write a business plan PPT for financing and then share the revenue from the financing.

But what we can do is to shorten the time it takes for customers to create a PPT from 2 hours to 2 minutes, and they are willing to pay for this result. Many corporate bosses recognize this. For example, the boss of Midea clearly stated that he would