The problem Negotiating Companion, deals find common ground solves

PWC launched the report of its consumer insights survey in feb 2023, which talked about the friction points of consumers in shopping experiences. So it surveyed 9000 consumers across 25 countries and still most preferred in-store shopping. It asked the next question to understand the feature importance in those in-store experiences. As I show you this graph, the role of sales associate was major. It is a dire need which helps businesses to manage customer emotions, and shape expectations. There is also a need for customer loyalty and conversion, which is highlighted in the survey. Many avoid online shopping because they can't haggle over prices and experience. Online marketplaces need to replicate this personalized experience to enhance user engagement and drive sales.

Platform facilitates online price negotiations, boosting customer engagement. Its unique offering attracts deal seekers, converting in-store shoppers to e-commerce. Rich negotiation data fuels innovative marketing strategies and industry advancements. It improves customer retention by addressing friction points and increasing deal conversions. It Drives e-commerce growth by enhancing sales through personalized negotiation experiences. Our AI salesperson as a chatbot acts as the product sales person and talks about the product & discusses related queries of the users. Now +user will prompt to request for the low prices for the product & also may offer their own prices. The chat bot will handle all these situations and try to close the deal as best as possible for the seller and describe the product features as well. Chatbot is bound with a minimum negotiable price below the threshold price it will not gonna accept or offer the deal. Making it optimal for both the user and the seller.

Challenges we ran into

One of the initial hurdles encountered was the complexity of using a generic AI model to generate accurate and contextually appropriate product prices. AI models excel at pattern recognition and text generation, but pricing requires a more structured approach that considers various factors. To tackle this challenge effectively, the project team developed a custom algorithm specifically tailored for price generation. This algorithm blended rule-based logic with machine learning techniques and incorporated historical pricing data, market trends, and competitor pricing to inform the price generation process. Iterative testing and refinement cycles ensured that the algorithm was accurate, relevant, and aligned with business objectives.

Deploying the AI model on Linux servers presented significant challenges, particularly related to compatibility issues and errors encountered during deployment on Azure Web Services. To address this, the project team adopted a strategic approach leveraging containerization and external deployment platforms. The backend application, including the AI model and related services, was encapsulated within a Docker container to ensure portability and encapsulation of dependencies. Deployment on OnRender—an external platform specializing in Docker container deployments—streamlined the process and mitigated compatibility challenges. Continuous monitoring, optimization, and thorough testing were conducted to verify stability and performance post-deployment.

Another challenge arose from the AI model's inability to generate human-like and engaging responses, particularly in simulating a salesperson's conversational style. To enhance the model's naturalness and engagement, the team focused on hyperparameter tuning and detailed prompt engineering. Extensive experimentation with different hyperparameter configurations optimized the model's performance, while crafting detailed prompts and input structures provided clear guidance to encourage more conte