Imagine entering an e-commerce live stream, captivated by a charismatic host, and finding yourself questioning whether that person is a genuine human or an artificial intelligence avatarThis uncertainty might be more common than you think as advanced AI technologies effortlessly integrate into online shopping experiencesApril 2024 marked a significant milestone in this domain when Liu Qiangdong, the founder and chairman of JD.com, made his debut as the AI digital persona "Cai Xiao Dong Ge," attracting more than 13 million viewers during the live broadcast—setting a new record for online viewership.

The rapid proliferation of AI digital humans within e-commerce platforms is transforming the landscape of online shoppingReports from JD.com indicate that their language-enabled digital humans, known as Yansai, have already been deployed in over 7,500 brand live streamsIn a move that showcases the transformative nature of AI, Yin Guandong, head of the Merchant Operations Department at Baidu E-commerce, disclosed that digital live-streaming generated an astonishing 11-fold increase in gross merchandise volume (GMV) over the past year.

Beyond the visible AI digital humans or virtual customer service agents that consumers interact with directly, artificial intelligence is gradually permeating various facets of e-commerce operations in 2024. This includes areas such as demand forecasting, content creation, personalized recommendations, intelligent search capabilities, and advertising strategy both in creation and distribution

A genuine synergy is emerging where AI models enhance merchant operations, driving efficiency and productivity.

According to Qiu Haoping, the AI chief at Weimeng Group, "We jokingly referred to the first half of 2024 as the era of 'agents having small meetings.' At that time, we were merely experimenting with one or more agents working in isolated segmentsBy the later half of the year, we achieved multiple agents operating together, resulting in efficiency improvements of up to 200% in certain segments." The reach of AI digital humans has expanded substantially, even reaching thousands of farmers.

Yao Huifeng is a long-time entrepreneur in the rice-selling business, having spent over a decade returning to his roots to sell riceHowever, efforts to sell online through platforms such as Taobao, Weidian, and Douyin yielded disappointing results, leading him to assemble five different e-commerce teams with each consisting of up to five members at their height

• Why H-Shares of State Banks Face Scrutiny
• The Stock Market Celebration is Coming to an End
• Rate Cuts on Hold Amid Strong Economic Data
• Lincoln China: 2024 Initiatives and 2025 Outlook
• Carbon Neutrality: An ESG Investment Case

However, the costs of operating a traditional live streaming studio proved too dauntingThe overhead to create a single live-stream session could reach around 1,200 yuan, plus monthly expenses exceeding 20,000 yuan due to staffingFaced with this financial burden, Yao was forced to dissolve his teams, retaining only a single customer service representative.

In contrast, setting up a digital human live stream has emerged as a more viable option for many online sellersUtilizing established digital human technology, Yao can now establish a new live streaming studio within five minutesAI-generated content for the live feed, including scripts and promotional materials, allows for continuous broadcasting 24/7, substantially reducing operational costsAs technological advancements lower costs, establishing a digital human studio became even more accessibleBy mid-2024, the price for a standard digital human dropped to below 10,000 yuan, while customized versions ranged from 20,000 to 80,000 yuan based on specific requirements.

Recognizing the transformative power of AI models, numerous established companies and startups flocked to this burgeoning sector

Notably, leading human-driven streaming organization Meione launched a trial of AI digital human live streaming during the 6.18 shopping festival, alongside intelligent customer service capabilitiesMajor e-commerce platforms like Taobao and Kuaishou are also experimenting with digital humans to enhance their shopping experiences.

Some platforms, aiming to promote digital human technology, offer merchants complimentary usage for a designated periodUnder government support, Zunyi Supply Chain E-commerce collaborated with Baidu to create over 100 digital human live streaming rooms, showcasing 1,000+ Guizhou agricultural products and generating sales exceeding one million yuan for over 300 local farmers.

JD.com highlights its progress, reporting that their Yansai digital humans now feature over 140 personalized personas and 50 industry-specific attributes, answering up to 90% of common inquiries with remarkable accuracy, surpassing 80% of human hosts in performance

"For small and medium-sized merchants without live streaming capabilities, our platform now enables a one-click creation of AI-generated live streaming rooms, effectively lowering entry barriers and increasing revenue," a representative from Baidu E-commerce explained, stating that using digital humans can reduce operational costs by over 80%, while GMV and conversion rates soar by 62% and 83%, respectively.

Wang Yunfeng, CTO of Zhidao Technology, pointed out that AI digital humans epitomize the capabilities of large-scale model systemsThey combine multimodal presentation capabilities with intelligence drawn from agent-based interactions and comprehensive service abilitiesIn 2024, digital humans exhibited significant advancements in mode representation, increasingly resembling real people with more vibrant expressionsBy January, JD.com’s Yansai digital humans had achieved advanced motion capture and choreography, transitioning from mere movement to a convincing semblance of life

This year also witnessed Baidu successfully reproducing real human live streaming with digital avatars leveraging standard 30-to-40-minute playback clips.

The traditional dual-host format is prevalent in human live streams, especially in specific product categories, proving to be more effective compared to single-host formatsDuring the Double 11 shopping event in 2024, JD.com realized a significant technological advancement by introducing dual digital human live streaming, officially available to merchants.

The period leading to December 2024 has seen thousands of farmers successfully utilizing digital humans for live commerceBaidu further announced plans to provide 200,000 Hui Bo Xing digital humans at no cost to assist merchants in their e-commerce operations.

In November 2024, Yao Huifeng began exploring the opportunity of operating a digital human live streaming room on JD.com, utilizing AI tools such as Yansai and AI customer services

With this support, Yao was able to independently manage a digital human streaming operation, showcasing the potential for individuals to harness AI technology effectivelyHowever, the landscape reveals discrepancies among platforms regarding digital human live streamingWhile Douyin, Kuaishou, and video account platforms lean towards promoting human hosts expressing concern over ecological impacts, JD.com and Baidu dedicate considerable resources to digitally powered live selling.

Moving beyond just digital avatars, the rapid evolution of AI capabilities within e-commerce operations in 2024 can also be seen in two major functional categories: generative content marketing and interactive marketingGenerative content marketing encompasses a variety of created assets from text-based articles to video productions, leveraging digital human representations as part of this diverse mix.

For instance, in digital human live streams, personalized and comprehensive content generation outputs scripts that are persona-driven and professionally tailored to a higher standard

Merchants can select from various script styles to suit their industry needs, thereby enabling a more customized promotional approach fueled by shared knowledge bases.

AI is streamlining the creation of product graphics and promotional materials as wellBy simply uploading an image of a product—without needing additional information like formulation or size—AI can extract crucial details to help craft catchy taglinesQiu Haoping remarked on the continued advancement of multimodal AI applications, stating, "While signs of progress appeared in 2023, by 2024, multimodal AI is widely applied within the e-commerce landscape." The improved precision of text-to-image generators and a lowered bar for image production highlight AI's growing potential in content creation.

Interactive marketing capabilities revolve around AI-driven customer service and conversational shopping assistants

JD.com’s shopping assistant, Yansai Guoguo, and Zhidao Technology’s assistant, Xiao Zhi, as well as Weimeng’s AI customer support demonstrate how these tools enhance user experience through multi-turn dialogues that intuitively guide customers toward desired products.

Wang Yunfeng elaborated on the functionality of AI shopping assistants, illustrating how these systems engage with users to pinpoint precise needs, ultimately suggesting the ideal products in a natural, personalized mannerFurthermore, Qiu Haoping noted that when merchants pose questions—such as “How should I configure my coupons?”—the AI not only responds directly but also assists users in navigating to specific areas for actionable suggestions.

However, turning the monumental capabilities of large-scale AI models into practical applications within the e-commerce space presents substantial challenges

The need for an authoritative reference (data), along with resolution around factual accuracy and real-time reliability, is crucial for bridging complex e-commerce operations where extensive knowledge and responsive data are integral to the success of AI.

In 2024, the widespread deployment of agents capable of integrating multiple tasks becomes instrumentalAgents are not simply limited to performing isolated functions; they can execute complex workflowsDuring the first half of 2024, Weimeng explored deploying agents for specific e-commerce functions like orchestrating promotional campaigns, ensuring comprehensive material production, and promoting unified execution.

By the later half of 2024, Weimeng expanded successful strategies across various e-commerce operations, exercising AI's capabilities in contextual settings while also striving for deeper engagement to address underlying user intentions and objectives.

The widespread embrace of AI live commerce is intrinsically tied to the adoption of agents

According to insights from Baidu, Hui Bo Xing represents an all-encompassing AI digital human live-streaming solution that boasts super-role capabilities, leveraging intelligent hosting, script automation, interaction features, and visual presentation that form an overarching intelligent system for live broadcasts.

As e-commerce transitions from its initial 1.0 phase focused on traffic to the 2.0 user-centric era, we now find ourselves on the brink of a 3.0 revolution—requiring merchants to adopt AI-centric perspectives in strategizing business logic and upgrading operational methodologies.

In the emerging intelligent e-commerce landscape, AI applications are poised to become the primary driver of incremental growthAs noted by Ping Xiaoli, Baidu's VP and Head of E-commerce, "With equal inputs, the more extensively AI technologies are applied, the quicker GMV accelerates and the greater the revenue potential." Subsequently, developments in 2024 spotlight two pivotal trends in AI applications within retail: cost reduction and the generation of new value streams.

"Within Zhidao Technology, AI's integration achieved impressive savings, particularly in community content development, suggesting that AI's efficiency predictions are achieving greater accuracy than human assessments," Wang Yunfeng added

"Recognizing that the deployment of AI for content generation is massively less resource-intensive further fuels this adaptation." In a glimpse of practical application, Weimeng found that when constructing marketplaces, AI’s contributions claimed approximately 70-80% of merchant management operations and nearly 90% for fulfillment configurations, drastically enhancing customer service efficiency by around 75%.

As an illustration of the burgeoning impact of this technology, JD's cloud branch reveals that over 800,000 merchants leverage the Yansai AIGC products, with 99.5% utilization for marketing text generation, yielding a conversion rate increase of 53% and a cost-saving exceeding 90% for video production.

Despite this momentum, it is essential to acknowledge that the evolution of large-scale AI models and their application in e-commerce remain in relative infancy

As industry leaders and stakeholders draw parallels to the early internet boom, they echo both excitement and caution regarding the rapid pace of innovation.

From a technological foundation perspective, a clear distinction is emergingWhile early 2024 saw discrepancies in AI model thresholds between domestic uses and international counterparts, substantial advancements were achieved by mid-year, heralding an upward trajectory configurationThe primary focus now revolves around optimizing data management processes—incorporating aspects like data cleansing, governance, indexing, retrieval, and matching model requirements.

Approaching innovation with large-scale models can feel like "walking across a river by feeling for stones," as Qiu Haoping articulatesThe challenge lies in capturing vague user requirements while keeping pace with swiftly evolving technologiesThere’s always the concern of launching applications that can outpace user adaptability, thus necessitating careful balance in transformative initiatives.

General consensus suggests that agents will significantly influence the trajectory of AI in e-commerce, particularly in enhancing operations regarding demographics, user experience, product analytics, and advertising effectiveness