Meesho elevates customer experience and boosts developer productivity with Microsoft Azure OpenAI Service

With a vision to enable 100 million small businesses to succeed online, Meesho is democratizing internet commerce and continuously facilitating MSMEs’ transition from the offline to the online realm. The company provides suppliers with a range of industry-first benefits such as zero commission and zero penalty. Over 15 lakh sellers are registered on Meesho, growing their business by tapping the company’s massive customer base, state-of-the-art tech infrastructure, pan-India logistics, and customer support capabilities.

With 80% of Meesho customers coming from tier 2 towns and beyond, the company has been instrumental in bringing new-to-e-commerce users online. Leveraging tech innovations, seller-friendly policies, and an asset-light structure, the platform has emerged as the lowest-cost destination catering to India’s price-conscious customers. Meesho is working towards enabling access and affordability for this set of consumers, reimagining the internet commerce journey for them.

Striving for excellence in customer service and developer productivity

Meesho manages a large contact-center team, and its transaction volumes surge significantly during sale events due to escalated demand, translating to thousands of chats daily. The leadership team at Meesho sought a scalable solution that could seamlessly handle the widely fluctuating volume of transactions without hiring additional support agents. They were aiming for a solution capable of measuring and augmenting developer efficiency and improving test coverage. At the same time, Meesho was also looking at optimizing operational efficiency across diverse facets of its business.

Recognizing the transformative potential of generative AI, Meesho’s leadership team saw it as a catalyst for improved productivity across diverse business functions. They decided to leverage generative AI’s power to develop more customer-centric solutions spanning the entire customer lifecycle. They aimed to infuse generative AI capabilities into the platform to personalize shopping interactions and elevate the overall customer experience. Generative AI-powered solutions could fuel the workforce's creativity and expedite innovation, enabling Meesho to deliver more customer delight and achieve greater productivity at lower costs.

Debdoot Mukherjee, Head of AI and Demand Engineering, Meesho, says, “We were looking for a generative AI solution primarily to turbocharge our customer service function. We also wanted to enhance our software development capabilities. Of course, it was important that these solutions integrate seamlessly with our existing IT landscape.”

Enhanced customer service, query resolution, and software development

Meesho began exploring various business use cases on OpenAI. Microsoft’s partnership with OpenAI in building scalable and cutting-edge AI platforms played a decisive role in Meesho's choice of a Microsoft solution. The team from Meesho participated in a workshop at the Microsoft Technology Center for Generative AI. It was an interactive session that included a hackathon for over 300 developers and collaboration between multiple teams from Meesho and Microsoft. Following this event, Azure OpenAI Service, GitHub Copilot, and vernacular language support for AI models were identified as key technologies to be used in the solution to be deployed.

Teams from Meesho and Microsoft worked together to deploy a multi-region, virtual chat agent based on Azure OpenAI Service and GPT-4. It is specifically designed to handle high-capacity chat conversations. The Microsoft team helped optimize prompts and implement a Retrieval Augmented Generation (RAG) pattern to reduce the overall token consumption per chat, thereby optimizing costs. On the development front, Microsoft introduced GitHub Copilot with Visual Studio Code for their developers and conducted an intensive three-day workshop to increase awareness.

The virtual chat agent went live across seven Azure regions and supported the sale events around the Dussehra and Diwali festivals. Meesho was able to handle 40% more traffic on customer service queries, and there was an impressive 25% increase in customer satisfaction scores. Meesho was also able to support 72 orders per second across top categories and it on-boarded 200,000 new sellers.

Following the close collaboration on daily and weekly reports on usage, recommendation adoption view and comfort index of its developers, Meesho adopted GitHub Copilot for over 200 developers to streamline code deployment processes and enhance unit testing.

Aiming to be future-ready with generative AI

As an online marketplace, Meesho aims to remain at the forefront of technological innovation, especially using generative AI. It is exploring use cases such as review summarization, merchant assistance, recommendations, and more. It also plans to gradually increase the deployment of the virtual chat agent during other large-scale sale events, such as the end-of-year sale, Maha (great) sale, and the Holi sale in the financial year 2024 (FY 2024). On the software development front, it is looking to integrate IntelliJ plugins for its GPT chat models.

Debdoot Mukherjee, Head of AI and Demand Engineering, Meesho, says, "The integration of generative AI capabilities in our systems has helped us manage customer queries more efficiently. We also witnessed a record 16 million app installations during our sale events. We are excited about the opportunities that generative AI can bring to our business. We look forward to continue collaborating with Microsoft and leveraging generative AI capabilities to do more with less."