Leveraging Generative Large Language Models for Optimizing Sales Arguments Creation: An Evaluation of GPT-4 Capabilities

A new era in Generative Artificial Intelligence has begun with the release of powerful Large Language Models (LLMs).These models have shown significant potential in marketing.Essentially, they have been redefining sales practices and changing sales funnel steps.Despite their promise and persuasive capabilities, there remains a dearth of comprehensive understanding regarding the capabilities of LLMs in generating sales arguments, which is a task traditionally characterized by high costs, time consumption, subjective biases, and the need for considerable expertise and skills.In this study, we take the first step towards exploring the ability of LLMs to generate sales arguments.To this end, we evaluated GPT-4 as one of the most capable LLMs to date in performing zero-shot sales argument creation from product features.We created a dataset containing textual descriptions of features extracted from brochures, catalogs, and technical data sheets of various products of a global company specializing in the manufacturing and retailing of furniture, appliances, and home accessories.We conducted a human evaluation with five experts, covering three main criteria of argument quality, namely coherence, persuasiveness, and relevance.The experimental results revealed the remarkable ability of GPT-4 to generate high-quality and well-structured sales arguments according to the Feature-Advantage-Benefit method.Over 98% of the evaluated arguments were coherent and persuasive.Regarding relevance, the model exhibited an accuracy of 91.53% to align the generated arguments to customers' purchase motives, namely security, vanity, novelty, comfort, money, and likability.These findings could lead to significant time and cost reductions, allowing the sales force to focus on higher-value tasks.We posit that this study heralds a novel avenue for exploring LLMs' capabilities in other steps of the sales funnel process.