2024-07-17
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
This is four or five times the growth of daily orders!
From placing 200 orders a day to continuous hot sales for a week, the owners of a domestic husband-and-wife startup store were inexplicably overwhelmed by the snowflake-like overseas orders.
This store is called Zeuslap. It sells auxiliary displays and does overseas business, especially in the Korean market.
On Black Friday of a certain year, we achieved the record of “over 1,000 orders in ten minutes”.
If you want to know the secret behind Zeuslap, the key point is that since November last year, the small team has started using AI to decorate their own stores.
Now, open Zeuslap overseas stores, banners, product scene pictures, product details pictures, as well as social media cover pictures such as YouTube and Tiktok——
The design and production of theseAI is the master。
Lu Xun once said (not really) that behind every merchant who uses AI to decorate their store, there is a team providing technical support.
The answer is revealed: Zeuslap decoration shop's AI image generation capability comes from a team that was established less than a year and a half ago.
Alibaba International’s AI team。
This team has a philosophy that the algorithm framework is important, the base model is important, but the most important thing is how many AI tools/capabilities are available for people to use.
The team revealed a set of data:
After using the platform's AI functions, merchants have seen improvements ranging from 1% to 30% in click-through rate, conversion rate, consumer satisfaction, cost reduction and other aspects.
Overseas e-commerce operations have become simpler
In fact, using AI to reduce the cost of e-commerce overseas expansion is not a new strategy in today's market environment.
In fact, for a small team like ZeuslapFor small and medium-sized businesses, with limited manpower and resources, AI can help them complete marketing planning in a more efficient and economical way.。
More energy can be invested in improving the quality of products and services.
Therefore, outsourcing marketing materials to AI has almost become a standard choice.
When browsing the Ali International AI overseas homepage, QuantumBit discovered thatImage generation is just the tip of the iceberg of its available AI capabilities.
"The value brought by AI is not simply reflected at the labor level," said Zhang Kaifu, vice president of Alibaba International Digital Business Group and head of AI business.
In other words, from product selection to after-sales service, many scenarios throughout the entire process have become simpler thanks to Alibaba International's AI technology.
Moreover, the more segmented the scenarios are, the more specific the efficiency improvement can be.
Here are a few~
Platform Customer Service Agent
When running an e-commerce business, no matter how self-service it is, customer service is indispensable.
The intelligent customer service built based on Alibaba International's AI capabilities can automatically convert English into 13 other languages, providing 24/7 multilingual customer service online (I, as a working person, am ashamed of myself).
In conversations with consumers, intelligent customer service can learn the details of various products and understand the semantics of the consumer conversation context to achieve a realistic and fluent response.
Moreover, such intelligent customer service has become standard for all fully-managed and semi-managed products on AliExpress, Alibaba's international cross-border e-commerce platform.
A few more words, most of the merchants behind full and semi-hosting are small and medium-sized enterprises, and they face problems such as language and time difference in pre-sales consultation.
The AI customer service, which has no language barriers and is always online, is really a great help.
Data shows that in June, the conversion rate of pre-sales inquiries for products equipped with AI customer service increased by 29%.
Chargeback Defense Agent
For cross-border merchants, the last thing they want to face is after-sales disputes, and small and medium-sized merchants are even more troubled by this situation.
Take Chargeback as an example.
Chargeback means that the cardholder refuses to pay the card-issuing payment institution for some reason after the transaction.
The international average refusal period is between 180 and 540 days. The reasons may be credit card theft, simple cargo disputes, or malicious refusal. According to statistics, the average cross-border e-commerce customer has been refused payment at least once in the past year, resulting in losses of more than US$89 billion.
How was the old solution solved?
When encountering a malicious chargeback, merchants need to spend 20 minutes (average time) to fill out the appeal materials, and then wait for the platform staff to review, or provide additional information.
Because it requires timely response and involves legal and financial knowledge, merchants always have a headache in this area.
Now facing this situation, merchants can use the Chargeback Agent launched by Alibaba International.
With Chargeback Agent, merchants can download the rebuttal template in the background with one click, and AI will automatically fill in the required information in the word document on their behalf.
It is generated in one second, and merchants do not need to fill it out manually.
Over the course of a year, the coarse cloth is estimated to help businesses recover losses of tens of millions of yuan.
Smart Refund Agent
In recent times, "refund" has become the most hotly debated topic in the e-commerce field, both in domestic and foreign trade.
The Refund Agent launched by Alibaba International AI team is extremely practical for this scenario.
Especially because unlike domestic trade merchants, when selling goods across borders, if consumers ask for refunds or returns without reason, the cost of shipping the goods back is very high.
Sometimes it’s even higher than the original price of the product…
However, if the merchant only refunds the money and does not recycle the goods, he will bear great losses in the goods.
Alibaba International's Return Agent uses AI to share the manpower that merchants should have borne, negotiate partial refunds and no-return policies, thereby reducing the return and refund ratio and helping merchants recover losses.
To elaborate, the return agent uses multimodal technology to complete tasks such as identifying disputes and recognizing vouchers, and then provides a refund plan for the non-returned part, and then calculates the refund amount.
In a nutshell, it can achieve "maximizing consumer satisfaction" and "minimizing merchant costs".
Of course, the above are just a few examples. In fact, Alibaba International AI can provide AI services in more than 40 scenarios.
"Both scenarios and models are important, but the process of finding scenarios is actually very empirical." Zhang Kaifu added, explaining why these scenarios were chosen for AI capability adaptation. "Regarding the priority of scenarios, our team has a relatively scientific calculation model, and will calculate the underlying data more strictly; but business value is not the only thing, and the boundaries of the model must also be considered. The R&D team and the technical team must keep up."
AI capabilities behind the scenes
Every leap in technology will lead to a wave of upgrading and replacement of existing industries.
Take the AIGC raw image, which was the first to make a breakthrough in this round of AI wave:
In August 2022, Stable Diffusion was open source.Leading effect and low computing power requirements have kicked off the commercialization of AIGC.。
But Stable Diffusion is not a panacea, especially there are bugs such as low controllability and high threshold for prompt words.
Later, technology practitioners launched technologies such as ControlNet, IP-Adapter, and LoRA one after another, just like patching.
△ControlNet
The development of AIGC raw images makes it possible to realize large-scale application of AI in the e-commerce field.
In March this year, Zhang Kuo, President of Alibaba International Station, said at an event:
AI is driving the foreign trade industry towards comprehensive intelligence, which will lead to a comprehensive reduction in foreign trade thresholds and an exponential increase in efficiency.
At that time, exactly one year had passed since Alibaba International set up its AI team for low-key testing.
In one year, Alibaba InternationalWe have been able to achieve large-scale application of AI capabilities in the field of cross-border e-commerceAmong them, in half a year, the number of products released by AI has reached one million, and through AI optimization, the overseas search volume of these products has increased by 37%.
Alibaba's international AI team now has more than 100 members. The team emphasizes that it is not a team that trains basic models. In the process of providing AI support, it focuses on three things.
That is, establish a multi-language enhanced model MarcoPolo, establish a multi-modal model MarcoPolo-VL, and provide MaaS (Model as a Service).
So, classified by type, what AI technology does Alibaba International provide to support the huge volume of calls?
The first is multilingual text generation technology.
One of the major difficulties faced by cross-border e-commerce is language and cultural barriers.
The multilingual text generation technology provided by Alibaba's international AI team can adapt product details descriptions to local languages, and also allow AI to rewrite and optimize multilingual titles for product introductions.
The second is AI image processing.
For example, in dealing with the frequently used problem of "creating e-commerce images from scratch", Alibaba International provides virtual try-on API technology.
Take clothing product images as an example. After the merchant authorizes, AI uses image pre-recognition capabilities to screen products that need to be virtually tried on, and then uses clothing segmentation capabilities to cut out the clothing that needs to be worn from the main product image with a complex background.
Generate multiple virtual fitting renderings with one click - currently a single image can be generated in 8 seconds, and is suitable for different models to try on clothes.
Next, you can select available images to replace them on the platform side.
It is worth mentioning that when it is necessary to generate product display pictures with real people, the technical level will call on small models that are more professional in the field to enhance the controllability and usability of the pictures.
andSuch AI processing flow can be replicated to all product categories.
It is understood that image design capabilities such as smart elimination and smart cutout are also gradually being opened up to merchants for use.
In short, Alibaba International AI technology can help with every aspect of store decoration.
Secondly, the multimodal large model provides recognition capabilities.
The team proposed and used a new multimodal large model architecture called Ovis (click here to learn more about Ovis).
Ovis draws on the text embedding strategy in large models and introduces a learnable visual embedding table, which first converts continuous visual features into probabilistic visual tokens, and then obtains structured visual embedding through multiple index weighting of the visual embedding table.
It has excellent performance in various multimodal tasks such as visual perception and life scenes.
For cross-border merchants, consumers have no reason to return goods for refunds, the cost of shipping goods back to the country is very high, and if the goods cannot be returned, there will be financial losses.
Therefore, Alibaba International AI provides an intelligent refund agent.
After the consumer applies for a refund and uploads the receipt, the Agent begins to use multimodal recognition technology to identify the reason for the dispute and analyze and verify the text/image/video receipt.
It then determines whether a refund is possible, intelligently determines the refund amount, and provides consumers with a partial refund if they do not return the goods.
Zhang Kaifu introduced that this capability has largely alleviated the lose-lose situation for merchants and consumers, and can save merchants 20 million in costs per year.
And there is reinforcement learning.
Through reinforcement learning, Alibaba International AI will score the products selected by the big model and predict the effects of the AI-generated materials.
In this way, the pictures put out are all pictures with good effects, ensuring the final delivery effect.
Of course, it is worth noting thatNone of the above AI technologies operate independently.
For example, in the advertising scenario, reinforcement learning is just one of many technologies.
This scenario uses a combination of AI capabilities such as multimodal recognition + multi-language text generation + AI image processing + reinforcement learning of delivery effects, and gradually accumulates an AI intelligent full-link advertising creative material generation solution for automated product selection, selling point extraction, and image generation.
Ultimately, the goal is to reduce material production costs and achieve better delivery results.
AI big model and overseas e-commerce, perfect adaptation
For merchants doing overseas business, the work has several more hurdles than domestic trade.
Although it seems diverse, if you look closely you will find that the many capabilities provided by the AI big model are the perfect fit for these scenarios.
In response to such market demands, which is also the inevitable result of technological development entering into practical application, Alibaba International established an AI team in April last year.
With years of experience in e-commerce and overseas expansion, we use AI to redo every step of cross-border e-commerce.
To put it another way, if a newbie merchant wants to do cross-border business, he or she will have to go through a lot of things, including product selection, store decoration, marketing, pre-sales shopping guide, and after-sales service...Every link is supported by Alibaba International’s AI technology.
It can be said that the rise of AI technology has brought revolutionary changes to overseas e-commerce merchants.
It not only improves the working methods and production efficiency of merchants, but also greatly enhances their competitiveness. The biggest beneficiaries are small and micro merchants who have greater room for improvement in their capabilities and energy.
It is conceivable that in the future, with the help of AI big model technology, e-commerce going overseas will develop in the direction of small teams, intelligence, personalization, and sustainability.
The Alibaba International Platform has become a(Especially small and medium-sized businesses)An important platform for embracing AI technology.
During this period, small and medium-sized businesses not only have the choice of easy-to-use and user-friendly AI technology, but they can also leverage Alibaba International’s deep experience and knowledge in the specific field of e-commerce/overseas business over the years - from a certain perspective, this just expands and practices the application scenarios of new technologies in the industry.
Looking at the current "general trend", at least two points are very clear:
First, the AI big modeltechnological change and ecological development,First, cross-border e-commerceThe global voyage.
Now, the gradual penetration and integration of the two has become a new feature of the era.
Both the merchant side and the technical side are facing new opportunities and challenges. Embracing AI big model technology can, to a certain extent, help cross-border e-commerce companies ride the wave of AI 2.0 and achieve global development and supply.
Now, Alibaba's international AI technology, which is at the forefront of the trend, has expanded the service objects of domestic AI big model technology achievements from the previous market of 1.4 billion people to a market of 8 billion people.
According to Zhang Kaifu, Alibaba's international AI capabilities have covered more than 40 application scenarios including marketing, customer service, product launch, design, and compliance, serving 500,000 merchants worldwide.
Further data shows that Alibaba International’s AI capabilities are called more than 50 million times per day, doubling in size every two months; a total of more than 100 million product information has been optimized.
Ali has a saying that has been widely circulated, namely“Make it easy to do business in the world”。
It turns out that you are not deceiving me.
Collect and organize the latest news information and make it available for free for everyone to read
