AWS for Industries
Lotte Homeshopping Reduces Human Agent Workload by 40% with Sendbird on AWS
Lotte Homeshopping is a major South Korean retailer that was faced with an increasingly complex quality assurance (QA) process. Their fast-paced fashion division struggled with product approvals that required manual coordination across multiple channels. Partner questions went unanswered for too long, while QA agents struggled to keep up with repetitive inquiries. These mounting challenges created significant delays in product launches and hampered Lotte’s ability to respond to market trends.
Perspective on the past
Lotte Homeshopping’s strength lies in its ability to source popular products and deliver them to customers through engaging broadcasts. However, to ensure consistent broadcast productions, products must meet specific compliance standards. As the number of products has increased, the compliance QA process has become a bottleneck. This has affected the fashion division, as seasonal changes and an increased number of suppliers multiply the work the QA team must handle. This has resulted in overworked staff and an increased likelihood that something could be missed in the QA process.
Meet Moni
To address these challenges, Lotte Homeshopping developed an AI agent named Moni. This solution was built using Sendbird’s AI agent platform, powered by Anthropic’s Claude 3.5 Sonnet large language model (LLM) delivered by Amazon Bedrock on AWS. Moni stands out for its accurate QA support and streamlined partner collaboration capabilities. The platform offers robust scalability for future AI services, comprehensive analytics for monitoring performance, and enterprise-grade security features that make implementation seamless within Lotte’s existing infrastructure.
The path to production: Rigorous testing
Before Moni could go live, Oh Ju-Young and the AI Promotion Team at Lotte Homeshopping conducted extensive testing to ensure the solution would meet their exacting standards.
“We conducted tests and POCs with several chatbot providers and were quite disappointed,” Ju-Young explains. “Sendbird’s accuracy and flexibility, paired with Claude, solved the problems others couldn’t.”
The team established a comprehensive testing protocol that spanned multiple dimensions:
- Accuracy testing across thousands of QA scenarios
- Security compliance verification
- Partner access management validation
- Response consistency checks
- Integration testing with existing systems.
“Unlike other AI services [we evaluated],” Ju-Young notes, “Sendbird comes with built-in features like two-factor authentication and message filtering by default, allowing us to seamlessly implement the service while meeting the high-security standards required by a large enterprise like ours.”
This rigorous testing phase, though time-consuming, proved crucial in selecting a solution that could handle the nuanced requirements of Lotte’s QA process. The combination of Sendbird’s platform and Anthropic’s Claude LLM provided the accuracy and reliability that previous solutions had failed to deliver.
What Sendbird and Lotte accomplished
Lotte Homeshopping’s journey began with six months of testing various solutions. Once they discovered Sendbird AI, the project accelerated rapidly. Ju-Young was able to complete a proof of concept, implement the solution, and successfully onboard their fashion partners very quickly. The integration of Sendbird’s user-friendly AI agent user interface with Anthropic Claude’s reliable LLM capabilities created a powerful solution that was both effective and easy to deploy.
Results
The impact of Moni became evident within just four months of its launch. Human agent workload decreased by 30-40% as the AI system handled routine inquiries efficiently. Partners worldwide could take advantage of around-the-clock support, while product time-to-market significantly improved. Both partners and employees reported higher satisfaction levels, leading to the expansion of Moni’s role beyond QA into areas such as internal IT support.
Conclusion
Lotte Homeshopping’s implementation of AI-driven customer service has set a new standard for enterprise AI applications. The company continues to explore new ways to enhance customer experience using various LLMs and AI models.
If your team is struggling with outdated ticketing systems, repetitive requests, and poor customer experiences, consider Sendbird’s AI Agent Platform.
About Lotte
With revenues exceeding $70 billion, Lotte ranks among Asia’s largest conglomerates. Lotte spans industries with operations in retail, food, hospitality, theme parks, and chemicals. Lotte’s expansive portfolio encompasses more than 90 business units across more than 30 countries. Its diverse operations include Lotte Department Store, Lotte Duty Free, Lotte Hotels & Resorts, and Lotte World, a major theme park in Seoul.
Lotte Homeshopping is the group’s televised e-commerce arm, and can be compared to QVC or HSN, but on a national scale, with a digital-first twist.
Lotte Homeshopping highlights products using several channels, including mobile, television, and ecommerce. They curate, verify, and launch new fashion and lifestyle goods daily, with new products broadcast under a 24-48 hour cycle.
AWS Partner Spotlight
Sendbird is the world’s largest private in-app conversations platform. Sendbird believes that conversations are at the heart of building relationships and getting things done. As such, they have built the world’s most proven conversations platform for mobile apps across chat, voice, and video. Industry leaders like Lotte, DoorDash, Reddit, and Paytm use Sendbird to drive increased transactions and loyalty for hundreds of millions of users every month. Relying on Amazon Bedrock for large language model (LLM) selection, such as Anthropic Claude, Sendbird offers the world’s most proven conversations platform trusted by thousands of companies.