Investigation of various elements of negotiation, including emotions, power dynamics, first offers, competition vs. cooperation, have contributed greatly to the understanding of the crucial and common interaction of negotiation. However, there remains a gap between many hypotheses theorized by current negotiation and empirical evidence drawn from a large sample size. Thus, observing the rise of AI technology in commercial and academic settings, our research introduces BizzyBots, an AI-powered negotiation platform, to explore the impact of various directed behaviors (single time or continuous) on negotiation outcomes. To demonstrate the potential of integrating AI chatbot into negotiation experiments, we investigate how different humor types (e.g., self-deprecating, sarcastic, and affiliative) and their timing and frequency influence negotiation dynamics, including immediate agreements and long-term relationship building. Additionally, we examine how varying concession patterns, such as large, small, matching, and reverse concessions, elicit concessions from counterparts and affect the negotiation’s financial and emotional success. Across four pilot studies involving 191 participants, we find that frequent violation of expected behaviors often elicited negative reaction from counterpart, resulting in generally smaller concessions, higher impasses, and lower (Economic Relevance of Relational Outcomes) scores. Yet, occasional (non-aggressive) violations with reasoned explanation accompanying offers showed increase in financial benefits and ERRO scores. We acknowledge the limitations of our pilot studies as they are still in the starting stage of more forthcoming investigations, but our use of the BizzyBots platform offers many possibilities for future research across many sub disciplines within the realm of negotiation  with increased efficiency and reduced (human or physical) resources required.