Reproduce “Learning When to Quit in Sales Conversations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Emaad Manzoor, Eva Ascarza, Oded Netzer
Researchers developed an AI "stopping agent" that leverages a fine-tuned generative language model to determine when to terminate unproductive sales conversations. This approach reduced total call time by up to 54% while increasing expected sales by up to 37% in an empirical study of outbound sales calls, highlighting inefficiencies in human dynamic decision-making.
Key insights
- Early linguistic cues within sales conversations can accurately predict call failure (e.g., GPT-4.1 showed 94% AUC after 60 seconds), yet human salespeople under-react to these signals.
- Human salespeople exhibit cognitive limitations in dynamic decision-making, often relying on salient but late-appearing cues and failing to integrate risk information effectively under time pressure.
- Imitation learning provides a stable and cost-effective method for solving optimal stopping problems with high-dimensional textual states, outperforming traditional DRL and classifier-based approaches.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.