Injecting Salesperson's Dialogue Strategies in Large Language Models with Chain-of-Thought Reasoning

TL;DR

The paper tackles the gap between task-oriented and open-domain dialogues by introducing SalesBot 2.0, a dataset enriched with coherent chit-chat, smooth intent-guided transitions, and robust transition boundaries. It further proposes SalesAgent, a chain-of-thought–driven model that jointly performs intent detection, policy selection, and response generation to control dialogue strategies in LLMs. Empirical results from turn- and dialogue-level evaluations across diverse user simulations show improved naturalness, coherence, consistency, and reduced aggressiveness compared with SalesBot 1.0 baselines. The work demonstrates the practical impact of integrating commonsense knowledge and CoT reasoning to build controllable, explainable sales-focused conversational agents, while noting reliance on prompt quality and LLM capabilities as limitations. Overall, this approach provides a scalable path toward more realistic and user-adaptive sales dialogues in real-world applications.

Abstract

Recent research in dialogue systems and corpora has focused on two main categories: task-oriented (TOD) and open-domain (chit-chat) dialogues. TOD systems help users accomplish specific tasks, while open-domain systems aim to create engaging conversations. However, in real-world scenarios, user intents are often revealed during interactions. A recent study introduced SalesBot, which simulates dialogues transitioning from chit-chat to task-oriented scenarios to train sales agents. Unfortunately, the initial data lacked smooth transitions and coherent long-turn dialogues, resulting in poor naturalness in sales-customer interactions. To address these issues, this paper presents SalesBot 2.0, an improved dataset. It leverages commonsense knowledge from large language models (LLMs) through strategic prompting. Additionally, we introduce a novel model called SalesAgent, trained on salesperson's interactions, using chain-of-thought (CoT) reasoning. This model excels in transitioning topics, understanding user intents, and selecting appropriate strategies. Experiments using diverse user simulations validate the effectiveness of our method in controlling dialogue strategies in LLMs. Furthermore, SalesBot 2.0 enhances coherence and reduces aggression, facilitating better model learning for sales-customer interactions.

Figures (4)

• Figure 1: Illustration of our proposed pipeline utilizing LLMs to generate human-like dialogues.
• Figure 2: Understanding results and policy strategies for SalesAgent
• Figure 3: Matched policy selected by models tuned on SalesBot 2.0 and 1.0.
• Figure 4: Ground Truth vs. Outputs