Do you know that Liang Xiaoqing, who was once popular all over the Internet and was known as the first sister of middle-aged and elderly clothing, is actually a post-90s generation. In a variety show, she transformed 116 Pose in 30 seconds, which amazed many viewers. For an e-commerce platform, the quality of the model's photos may directly affect the sales of the product. In fact, many models who take photos for e-commerce platforms work very hard, and it is said that there is a record of 485 consecutive photos of clothes in one day, one stop for a day, and there is no time to even eat.
Now the advent of generative AI may give these busy models a breather. "With the help of generative AI, we have unlocked a new way of creation, that is, using our self-developed GenAI marketing material creation platform, which can achieve minute-level image production and improve the efficiency of material production by 30%." Dr. Zhongliu Xie, Chief Scientist of Shoplazza, further explained, "Based on the GenAI marketing creative platform, combined with Amazon SageMaker, Amazon EKS, Amazon ECR and other products and services, and working seamlessly with Amazon S3, Amazon EMR, Amazon Redshift, etc., we are able to quickly and mass-produce a variety of images in different styles. It avoids time-consuming and laborious work such as live model fitting and background replacement, and is also more conducive to the generation of some popular product ideas, which not only efficiently meets the needs of different businesses for material image processing, but also greatly saves the time and cost of product development. ”
"To do cross-border e-commerce business in a new way, we must now actively and closely embrace generative AI." Xie Zhongliu said, "A new paradigm of cross-border e-commerce in the era of generative AI has begun. ”
Embrace generative AI now
Generative AI and large models have been popular for more than a year, and "AI + e-commerce" has quickly entered the application maturity stage from the initial hype stage. Generative AI and large models can not only provide more efficient and practical tools for e-commerce, but also play a central role in many scenarios such as intelligent product selection, AI shopping guide, intelligent marketing, and intelligent customer service, helping cross-border e-commerce reduce content production costs, and are likely to reshape the development pattern of the e-commerce industry in the future.
As a leading SaaS platform for e-commerce independent websites in China, Shoplazza Technology is actively exploring the application of generative AI technology in various complex scenarios of cross-border e-commerce, including search recommendation, intelligent customer service, material generation, AI website building, etc., hoping to bring different application experiences to customers and even change the operation mode of the entire e-commerce.
"From a technical point of view, AI itself is a relatively simple and simple concept, but its impact on applications is huge. From a business development perspective, if e-commerce companies do not embrace generative AI today, it will be difficult to gain a foothold in the industry competition. Xie Zhongliu said, "It turns out that e-commerce business often requires a large number of operation personnel, customer service personnel and material production personnel. But now, generative AI-enabled search recommendation, intelligent customer service, material generation, AI website building, etc., have greatly changed the operation mode of cross-border e-commerce business, and intelligence, automation, simplification, high efficiency, and low cost have become the main theme. In the future, as generative AI technology continues to mature, this trend will be further strengthened, and it may only take a few people to take on the entire e-commerce operation. ”
In fact, this kind of change is not uncommon in other industries and fields. Last year, when generative AI was just emerging, a number of eye-catching videos appeared on Bilibili and Xiaohongshu, and it was said that behind them were individuals who continuously generated content through AI and made a lot of profits. Looking at the essence through the phenomenon, whether it is in the Internet, in the field of e-commerce or even in more traditional industries, generative AI will bring about a change in direction. It will not only change the mode of production, business model, and work efficiency, but also profoundly change the entire industrial ecology.
The four major scenarios of generative AI are the first to land
Shoplazza serves more than 360,000 cross-border e-commerce customers in more than 150 countries and regions around the world, covering a variety of cross-border e-commerce groups such as cross-border whites, boutique sellers, non-source sellers, TikTok sellers, industry and trade integration, brand merchants, foreign trade factories, Douyin sellers, Amazon sellers, and domestic platform sellers. Shoplazza provides a variety of convenient features, including high-conversion store theme templates, drag-and-drop theme editors, one-stop payment solutions, efficient multi-process store management, multi-platform advertising tracking and optimization, etc., to help customers quickly build websites, thereby accelerating the brand's overseas expansion and supporting cross-border e-commerce operations.
At present, in terms of generative AI, Shoplazza focuses on four major scenarios, including "GenAl marketing material creation platform, AI website building Copilot, intelligent customer service, and intelligent recommendation & search".
"The application of AI in marketing material creation is very promising, and it may become a new business growth point in the future." Xie Zhongliu said. In terms of GenAl's marketing material creation platform, because the e-commerce scenario involves a large number of marketing creations, including various marketing pictures, videos and copywriting, manual creation is time-consuming, laborious and costly, requiring investment in models, photographers, makeup artists, designers, venue fees, transportation expenses, etc., and the creation cycle can be as short as a few days or as long as weeks. Using GenAl marketing material creation platform, not only does not require high investment, but also can achieve minute-level image production, and at the same time provide a variety of style options, and create multiple photos at one time; At the same time, it covers multiple usage scenarios and covers different product categories, including virtual model fitting, product background replacement, and creative variation of popular products.
According to Xie Zhongliu, behind the GenAl marketing material creation platform, Shoplazza uses services such as Amazon SageMaker and Amazon ECR SD mirroring. For example, when developing a creative feature, Shoplazza optimized it by using Stable Diffusion XL1.0 in Amazon SageMaker JumpStart to experience the latest machine learning trends and models with one click. At the same time, Amazon SageMaker JumpStart also provides responsible AI technology, which has its own functions such as pornography detection and fear detection, and if the output generated images involve pornography, the Stable Diffusion XL1.0 model supported by Amazon SageMaker JumpStart will automatically code these images to avoid risks.
By using Amazon SageMaker's fully managed services, as well as efficient distributed computing and automatic parameter tuning, Shoplazza has greatly improved the speed and efficiency of machine learning model training, shortened the product development cycle from the past 6 months to 2 months, and significantly reduced the time cost of model training and tuning.
Shoplazza is keenly aware of the need for many customers to build a one-stop website. Based on Amazon Bedrock, Shoplazza Technology has integrated into the Claude3 series model, which realizes dialogue interaction and makes it easier to understand customer needs. Based on the operational data of Amazon RDS, you can build an AI website building data base, which can quickly disassemble requirements, build elements, and use these elements to quickly build a website.
In many industries, intelligent customer service is the first scenario for AI to land, and practice has proved that its effect is very significant. Shoplazza's SaaS website building service has a large number of customer inquiries every day, of which 69% of historical repetitive questions are repeated, and 92% of the answers to these questions can be found in the relevant help documents. In order to improve efficiency and reduce the workload of personnel, Shoplazza built RAG (Data Retrieval Enhancement) based on Amazon ElastiCache Redis cache vector data, and used the generated Q&A data to answer new similar questions. At the same time, based on Amazon Bedrock, the Claude 3 model is connected to build intelligent AI customer service. By applying generative AI to customer service, Shoplazza has reduced the average time it takes to resolve customer issues by 70% and customer service costs by 40%.
The application of generative AI to intelligent recommendation and search has brought new benefits to Shoplazza. In general, in the field of intelligent recommendation and search in cross-border e-commerce scenarios, the main issues to consider include how to unify multilingual search, how to understand the attributes of products, how to do multimodal matching between search terms and products, and how to do efficient personalized display and sorting. Shoplazza Technology uses Amazon Bedrock to access the Claude 3 model and applies it to technical scenarios such as content understanding, search intent judgment, and multi-modal product matching and recall, thereby improving the effect of intelligent product sorting.
"Search and recommendation are relatively traditional scenarios in the e-commerce industry. In this regard, we are technically iterative. Large models enable better feature extraction and better matching. "Starting from March 2024, Shoplazza Technology has made an initial attempt on the Claude 3 model. Prior to this, we have used many models, including the Mistral series model, as well as Meta's Llama series model and ChatGPT model. When it comes to model replacement, it is imperceptible to the customer. We test to determine which model is more cost-effective than one, and we use multiple models in parallel over a period of time. We can't predict what better models will be available in the future. Therefore, our basic strategy is to constantly follow the latest model and technology changes to seek better business opportunities. ”
Talking about the issue of AI security and compliance that cannot be ignored, Xie Zhongliu said: "This problem is really a headache. As the technology matures, we will continue to optimize our products and practices to better meet the security and compliance challenges that come with it. For example, we will use relevant technologies and tools to detect and post-process typical pornography and violence-related issues, and then present them to customers to minimize risks. In short, we will do everything possible to find the ideal solution in accordance with the existing laws and regulations, combined with relevant business practices, to escort the business development of our customers. ”
Work together to build a solid data foundation
Shoplazza's innovative application in the four major scenarios of generative AI relies on the support of a strong data base at the bottom. For example, Amazon S3 is used to store massive user data and marketing materials for model training. Leverage Amazon EMR Flink to run distributed data tasks to achieve real-time data cleaning and processing. Amazon OpenSearch is used to store vector data to achieve ultra-fast product search. Improve user rights management with Amazon IAM. Amazon Web Services' data base supports the operation of Shoplazza's multiple AI products, saving it 40% of customer service costs, improving material production efficiency by 30%, and increasing search and recommendation revenue by 20%.
Judging from the successful practice of Shoplazza, the landing application of generative AI is inseparable from a comprehensive, powerful, secure, and reliable data foundation. This is also a premise and foundation for the close cooperation between Shoplazza and Amazon Web Services for many years. Today, all of Shoplazza's SaaS website building services have been migrated to Amazon Web Services, and compared with its earlier systems, it has greatly improved system performance, response speed, cost control, data stability and security, and can provide customers with 99.9% availability. Amazon Web Services has abundant global resources and a complete service system, which can effectively reduce the burden of internal infrastructure operation and maintenance of enterprises, save time and energy in infrastructure management, and allow its technical personnel to focus on business innovation.
In particular, Amazon Web Services has complete functions, tools, and solutions for security protection and compliance, built a complete set of cloud security systems, and played an important role in Shoplazza's PCI level 1 compliance certification. In addition, Amazon Web Services' Cloud Initiative not only provides resource support for Shoplazza, but Amazon Web Services partners also provide very professional migration services, and the amount of terabytes of data is completely insensitive during the migration process.
"Amazon Web Services' in-depth insights into generative AI application scenarios in the e-commerce industry enable us to provide customers with a better user experience, achieve higher output efficiency, and have lower production costs." "In the future, Shoplazza will continue to optimize the GenAI marketing material creation platform, including improving the accuracy of the model, expanding the functions and adapting to a wider range of application scenarios. At the same time, we are also trying to use generative AI products related to Amazon Web Services large language models to achieve application scenarios such as intelligent question answering. We hope to continue to work hand in hand with Amazon Web Services to deeply integrate generative AI technology and the needs of e-commerce business scenarios, and help merchants achieve more innovation and business growth in the e-commerce field." ”