Original · Unique Research · 2026-05-13
Editor's note: The first-person interview and its judgments belong to the original Chinese author. This English rendition retains the opening essay, all interview sections, and the complete 10-question Q&A. Founder background, company history, market claims, and technical terms are source or speaker claims, not independently audited findings. Company, personal and work titles are transliterated where official English forms remain unverified. The term "vibe coding" is retained as the speaker's wording.
Unique Awards · Guest Interview
In Late 2021, Chen Guang (Billy) Was Doing Something That Seemed a Bit Ahead of Its Time
He was betting not on Amazon sellers, not on bulk-product going global, but on "brand going global."
"
There is a whole era of difference between Chinese brands "going global" and Chinese goods "being sold overseas."
In late 2021, Chen Guang (Billy) was doing something that seemed a bit ahead of its time—he was betting not on Amazon sellers, not on bulk-product going global, but on "brand going global."
Back then, terms like "AI Agent" and "large model going global" did not have the heat they have today. Most discussions in cross-border circles were about product selection, testing items, and spending on traffic.
More than three years later, the judgments he made back then—"the direction is right but the market is not ready yet"—are being verified one after another.
He is qualified to say these things. Tsinghua MBA, former CEO of Ronglian Qimo, and he led the first domestically founded customer service SaaS company to list on a US stock exchange—these experiences stacked together gave him a foundational logic others did not have when he founded QuickCEP.
I chatted with him for over an hour before the Unique Awards in Shenzhen.
There Is a Whole Era of Difference Between Chinese Brands "Going Global" and Chinese Goods "Being Sold Overseas"
Billy said his initial driving force for starting QuickCEP was not complicated: he saw the inflection point of China moving from "selling goods going global" to "brand going global."
The two things look similar, but in reality they are worlds apart.
Selling goods means just getting things sold. Brand means making consumers remember you, trust you, and come back to you repeatedly. The former can be solved with product selection and supply chain; the latter requires managing relationships.
"The Day One we created is oriented toward the future, preparing for the operational form of brand going global." Behind this sentence lies a considerable bet.
Back then, most cross-border tool companies were still helping sellers manage inventory, customs declaration, and shipping. QuickCEP went straight for "consumer operations" from the start—there was little competition in this track at the time, but there were also few customers willing to pay for it.
He did not explain how difficult it was back then, but simply said: "According to our observations, the cross-border e-commerce industry is developing toward product innovation, brand operations, and user operations."
I think this sentence is worth pondering.
He did not say "changes are happening"—he said "it is developing in this direction." That means this change has not finished happening, but the direction is already clear. For people building SaaS, this is a key distinction: are you arriving late to the market, or are you waiting for the market to come to you?
Customer Service Is Just the Tip of the Iceberg
"Are you guys doing customer service?"
This is probably the question QuickCEP gets asked most often. It is also the question that leaves Billy most at a loss for whether to laugh or cry.
He explained to me the gap between what he understands as "consumer operations" and "customer service":
"We understand customer service as just one segment of consumer operations."
If you only do customer service, you solve the moment when "the consumer comes to you." But in a consumer's lifecycle, the moments when they actively reach out to you are only a small part. Most of the time, consumers are silent—they are scrolling videos on other platforms, searching for competitors on Google, hesitating about whether to make a repeat purchase.
These silent moments are the real battlefield of brand management.
Billy listed the core pillars of a brand clearly: "First, innovative and high-quality products. Second, excellent service experience. Third, good interactive relationships with consumers."
He said QuickCEP wants to solve the latter two.
This positioning means what they sell is not a customer service tool, but a consumer relationship management system. Help Desk, AI Chatbot, CDP (Consumer Data Platform), marketing automation—these modules look like independent products from the outside, but internally they are an organic whole, not sold separately.
"They are just one module within CEP."
What Is the Most Overlooked Growth Asset for Cross-Border Brands?
I asked him, among today's going-global brands, what is the most easily overlooked growth asset?
He did not hesitate and said directly: private-domain users, or first-party data assets.
"Not many brands do this well."
This judgment is not just his personal feeling. The entire industry is going through something: traffic is getting more expensive, platforms' natural dividends are shrinking, and advertising ROI is declining. Everyone knows where the problem is, but most people's response is to "switch to another platform and burn money," rather than "manage the existing user relationships well."
What are private-domain users? They are people who have already bought your product, interacted with you, left an email, or added you on WhatsApp. They are consumers who have had a real relationship with you, and they are a brand's most valuable raw material.
But in cross-border circles, this area has long been overlooked.
The reason is not complicated—private-domain construction is slow, costs are not显性 (显性/visible), and short-term ROI does not look good. Compared with directly buying traffic, private-domain operations need to be "nurtured," require patience, and also require tools.
For QuickCEP, a tool company, this market gap of "nobody taking it seriously" is precisely their opportunity.
What Is the Essential Difference Between an AI Agent and a Traditional Customer Service Bot?
"What is the biggest difference between your AI Agent and a traditional customer service bot?"
Billy's answer was quite technical, but I will try to translate it into language ordinary people can understand.
The logic of a traditional customer service bot is: you ask, it answers. If it has the answer, it checks the knowledge base; if not, it transfers to a human. It is passive, has no memory, does not know who you are, does not know what you bought last time, and does not know whether you have filed a complaint recently.
QuickCEP's current AI Agent is different.
"It is an Agent with subjective initiative"—this is Billy's exact wording. Subjective initiative means it does not wait for you to ask; it knows what action to take at what juncture.
He mentioned several keywords: a complete Harness shell (rich toolbox), MCP protocol, Skill system, complete consumer Memory, and integration with the brand's e-commerce system and ERP system.
The business meaning behind these technical details is: this Agent knows who you are, what you have bought, your order status, whether the issue you raised last time was resolved—it can directly help consumers solve problems, rather than kicking you to the next segment.
Of course, there is a practical issue he did not avoid:
"Whether an enterprise-grade AI product works well depends on the joint cooperation of both parties, and it is a process of continuous improvement."
He said the way to avoid "looking very smart but having insignificant business results" is not to bet the project's success or failure on a single instance of "whether it is smart," but to provide a soil— a product system—that allows AI effects to continuously iterate.
This is down-to-earth. I have seen too many enterprises deploy AI tools and just wait for them to automatically produce results, only to be disappointed in the end. In reality, AI implementation is more like a new employee onboarding—you have to feed it data, allocate resources, and do calibration before it gets better with use.
From Selling Tools to Selling Results: How Long Does This Transition Take?
Moving from "selling software" to "selling business results" is something the entire AI SaaS industry is talking about. I asked him whether this actually holds true in the cross-border e-commerce industry.
Billy's answer was more nuanced than I expected.
He said he strongly endorses this direction, "but this process does not happen overnight; it takes time to gradually develop into pay-by-result. Each segment is uneven."
Then he raised a very interesting point—besides features and results, there is an intermediate state: pay-by-compute.
"Right now, Claude and OpenAI's business models are pay-by-token, and I think this is more practical than pay-by-result."
He explained the reason: pay-by-result is often unfair to the vendor. Because business results are affected by many variables—industry fluctuations, the client's internal processes, the client's product strength—not something a SaaS company can decide alone. "Simply assessing by results is unfair in most scenarios."
Pay-by-compute, on the other hand, is a more direct intermediate solution: you pay for how much computing resource you use. This makes the pricing logic clearer and makes AI's value more quantifiable.
I find this judgment quite valuable as a reference. Right now many AI companies are eager to promote "pay only for results," but at the commercial implementation level, this model is actually not mature yet. In most scenarios, pay-by-compute will emerge first.
China AI Going Global: Advantage in Speed, Shortcoming in Sales
I asked him how he views the opportunities and shortcomings of Chinese AI SaaS going global.
The advantage, he said, is iteration speed and cost-performance. "In the vibe coding era, I think engineering capability has been leveled."
This statement is a bit subversive. In the past, everyone always felt that overseas teams had strong engineering capability, which was a barrier for Chinese companies. But vibe coding has lowered the barrier to code generation, and engineering capability is no longer a moat.
So the real competitive dimension becomes: who iterates faster, who has lower costs, and who understands business scenarios more deeply.
And the shortcoming? He was very direct: local sales and brand trust.
Not technology, not product—sales capability and brand recognition. These two things are hard to solve with algorithms; they require real people to go local, build teams, build relationships, and build trust.
This is the hardest hurdle for most Chinese SaaS companies going global. Products can iterate quickly, but local sales connections are accumulated slowly, and brand trust takes even more time to build up.
In Three Years, Who Will Emerge?
I finally asked him for his judgment on the next three years of AI going global.
He said AI changes too fast, and a three-year prediction does not mean much. But he offered one judgment he considers certain:
"Vertical AI applications with deep industry know-how are not easily swallowed by model vendors. This point is certain."
This sentence is well worth savoring.
The capabilities of underlying models are improving rapidly—OpenAI, Anthropic, Google will get stronger and stronger, and they will consume most of the AI capability in general scenarios. But those AI applications deeply embedded in specific industries—such as products deeply integrated with cross-border e-commerce ERP, logistics systems, and full consumer lifecycle data—are things model vendors will not specifically do in the short term.
This is the moat of vertical AI applications: not how strong the model is, but how deep the industry data is, how well the business scenarios are understood, and how solid the upstream and downstream integration is.
For going-global brands, Billy's advice is very direct: if you are a growth-stage brand, the highest priority right now is to fill in your user operations capability.
Not AI, not technology—the capability to manage consumer relationships.
Traffic can be bought, but user relationships cannot. Your private-domain users are the trump card that will not become invalid because of changes in platform algorithms.
Interview Highlights Q&A
Q: How do you define QuickCEP as a company?
Chen Guang Billy (QuickCEP CEO): A one-stop global consumer service and operations AI Agent platform built for brand enterprises.
Q: Why did you choose the brand-going-global track back then? At that time, not many people were talking about this.
Chen Guang Billy: We valued the opportunity of the rise of Chinese brands going global, that is, the era opportunity of China shifting from "selling goods going global" to "brand going global." The Day One we created was prepared for the future.
Q: You previously had experience in cloud customer service, SaaS, and cloud communications. What was the biggest help from this experience?
Chen Guang Billy: These are the basic skills for doing B2B. The basic skills must be solid, and then combined with new AI capabilities to build products. The aspects you mentioned have all been particularly helpful so far.
Q: What segment of cross-border e-commerce has been most deeply changed by AI?
Chen Guang Billy: So far, customer service, marketing content generation, and insights segments have progressed relatively quickly.
Q: In cross-border e-commerce, what is the most easily overlooked growth asset?
Chen Guang Billy: Private-domain users, or first-party data assets. Not many brands do this well.
Q: What is the biggest difference between your AI Agent and a traditional customer service bot?
Chen Guang Billy: In its current Agent form, it is an Agent with subjective initiative, with a complete Harness shell—such as a rich toolbox, MCP, Skill—with complete consumer Memory, and integrated with the brand's e-commerce system and ERP system. It can directly help consumers solve problems.
Q: Enterprises deploy AI tools but the results are not obvious—how to break through?
Chen Guang Billy: Whether an enterprise-grade AI product works well depends on the joint cooperation of both parties, and it is a process of continuous improvement. The way to avoid this problem is: provide a safe soil for continuously iterating AI effects, rather than relying solely on a single instance of "whether it is smart" to determine the project's success or failure.
Q: AI SaaS moving from selling features to selling results—what do you think of this change?
Chen Guang Billy: I strongly endorse this direction. But I think besides features and results, there is an intermediate state of pay-by-compute. Because the results of each segment are not something the vendor alone can decide, simply assessing by results is unfair in most scenarios; pay-by-compute is relatively fair.
Q: For Chinese AI SaaS going global, what are the biggest advantages and shortcomings respectively?
Chen Guang Billy: The advantage is iteration speed and cost-performance; in the vibe coding era, engineering capability has been leveled. The shortcoming is local sales and brand trust. There are no shortcuts for these two things.
Q: In three years, which type of AI company will emerge?
Chen Guang Billy: Vertical AI applications with deep industry know-how are not easily swallowed by model vendors. This point is certain.