Language agents for optimal stopping
Stopping agents are language agents — large language models that
generate decisions — specialized for optimal stopping of conversations.
Specifically, stopping agents observe the ongoing conversation text and
make sequential wait or quit decisions that optimally tradeoff between waiting
to accumulate more information and incurring waiting costs.
For more details on the underlying theory and reference implementation, check out our website at stoppingagents.com and our paper on learning when to quit in sales conversations.
A stopping agent for sales: As an example of something you can build with stopping agents, check out our demo app of stopping agents in action providing stopping advice during an outbound sales call.
Manzoor, Emaad, and Ascarza, Eva and Netzer, Oded. "Learning When to Quit in Sales Conversations." arXiv preprint arXiv:????.????? (2025).
We thank OpenAI for API credits and acknowledge financial support from the Cornell Atkinson Center for Sustainability and the U.S. National Science Foundation.