Conversational Solutions for the Sales Team by leveraging GenAI techniques

Gupta M, Goel A, Gupta D, kamra G, Kumar R

In today's rapidly evolving business landscape, the demand for innovative solutions to enhance efficiency and drive growth has never been more critical. Recognizing this imperative, in this project, the team leveraged advanced technology to streamline operations and boost sales. The rationale behind this initiative is not merely operational efficiency but a fundamental shift towards data-driven decision-making.

By integrating state-of-the-art tools and methodologies, the team aimed to optimize workflows, reduce operational bottlenecks, and ultimately, accelerate sales performance by creating a sophisticated conversational AI tool on AWS. This tool  empower the client’s Sales Team, including Field Associates, Managers, and Leads, to gain profound insights from their sales data.

The conversational solution thus developed, using Generative AI techniques to enabled the Sales Team extract valuable insights from sales data, improving their understanding of customer behaviour, order volumes and product mix. The team implemented the system to simplify dataset interactions and increase engagement among Sales Team members, enhancing overall performance and key performance indicators (KPIs) for more strategic decision-making. By doing so, the team significantly reduced the time taken from raising queries to obtaining desired outputs, facilitating swift and precise decision - making processes and also reliance on data analysts for routine queries, freeing them to focus on more complex analytical tasks to ensure efficient resource utilization within the organization.

The project's primary objective was to develop an advanced conversational solution using cutting-edge Generative AI techniques to enhance the Sales Team's ability to interpret intricate sales data. Leveraging AWS services such as Sagemaker, Lex, Lambda, CloudWatch, and API Gateway, along with powerful language models like FlanT5-xxl and NSQL-Code-Llama-2-7B, the goal was to create a conversational UI for text and voice queries.

The project initially focussed on single interactions, specifically singleton SQL queries on single tables, excluding complex queries. It developed end-to-end solutions encompassing Conversational UI, text-to-SQL, and data-to-insights functionalities as a Proof of Concept (POC) with essential test cases, with deployment deferred until post model fine-tuning.

The project operates on a per query basis, lacking continuity across conversations, supports only English inputs without image processing and the response time varies based on computational power. Despite these limitations, the innovative approach aims to transform Sales Team interactions laying a foundation for scalability. The ultimate goal of the project is to establish a more comprehensive and sophisticated conversational system.

The solution bridges gap between raw data and meaningful insights, empowering the client’s Sales Team with a powerful tool for strategic decision-making, enhanced user engagement, facilitates data-driven decision-making decreasing the time required for insight extraction. To top it all, reliance on human resources is reduced, freeing up valuable resources for more nuanced and strategic tasks, enriching overall workforce dynamics.

Several key objectives were thus achieved: developing models capable of capturing and linking ideas across multiple turns to generate SQL queries aligned with the user's overarching goals, fine-tuning models on joins for higher prediction accuracy in multi-table scenarios, and considering scalability to multiple languages beyond the current English-based solution.

The project enhanced the conversational AI tool for the Sales Team through careful model selection, prompt engineering, and user-focused recommendations, leading to improved accuracy and usability. This resulted in enhanced decision-making, user experience, efficiency, potential financial benefits, and alignment with broader goals, with suggested ongoing improvements for sustained positive impact on user value and business outcomes.