Persistent Retrieval for Conversational Agents: A Study of Retrieval Efficiency
Karthik Ravula and colleagues at ServiceNow compare three ways a conversational agent can retrieve enterprise data, and propose a persistent retrieval subagent that keeps its retrieval history across turns and returns raw evidence to the main agent instead of a written summary.
Ask this paper
Designs compared. Skill-based retrieval inside the main agent, a non-persistent subagent that writes a findings response, and a persistent subagent that returns evidence directly and accepts correction requests.
Against skills. Across five multi-turn settings, the persistent subagent uses 43.3% fewer agent-model tokens, costs 45.0% less and responds 9.1% faster, with comparable accuracy.
Against non-persistent. It is 15.6 seconds faster per question but uses 65.1% more tokens and costs 29.1% more.
Ablation. Removing direct evidence return raises response time and cost more than removing persistence.
Scope. Evaluated on structured enterprise records in a demonstration environment.
Abstract
Conversational agents can retrieve information directly or delegate retrieval to a subagent. This design choice affects how retrieval history is maintained, what context each model call receives, and what information must be re-retrieved in follow-up questions. We compare three designs: skill-based retrieval in the main agent, a non-persistent retrieval subagent that generates a findings response, and our persistent retrieval subagent, which maintains retrieval history across turns and returns retrieved evidence directly to the main agent. Our approach also allows the main agent to request corrections from the same subagent. We evaluate the three designs on enterprise questions posed independently, within shared conversations, and as dependent follow-ups. Across five multi-turn settings, our approach uses 43.3% fewer recorded agent-model tokens and incurs 45.0% lower estimated API cost than skill-based retrieval, while reducing average response time by 9.1%. Compared with non-persistent retrieval, it reduces response time by 15.6 seconds per question, while using 65.1% more tokens and incurring 29.1% higher estimated API cost. Our approach shifts the retrieval subagent from producing synthesized responses to maintaining retrieval state and supplying evidence that the main agent can directly evaluate and use for answer generation.