Agent WARPP: Workflow Adherence via Runtime Parallel Personalization

Large language models (LLMs) are increasingly applied in task-oriented dialogue (TOD) systems but often struggle with long, conditional workflows that involve external tool calls and depend on user-specific information. We present WARPP (Workflow Adherence via Runtime Parallel Personalization), a training-free, modular framework that combines multi-agent orchestration with runtime personalization to improve workflow adherence in LLM-based systems. WARPP dynamically prunes conditional branches based on user attributes, reducing reasoning overhead and narrowing tool selection at runtime. Implemented using the OpenAI Agents SDK, our system parallelizes a Personalizer agent alongside a team of modular agents to personalize execution paths across domains. We evaluate WARPP across five intents of varying complexity in three domains—banking, flights, and healthcare—using condition-rich synthetic datasets and LLM-driven user simulators. Compared to both a non-personalized multi-agent setup and a standard ReAct baseline, WARPP cuts average latency by roughly, shrinks context size (measured in total tokens), and sustains equivalent end-to-end accuracy without any additional training.