MISCON: A Mission-Driven Conversational Consultant for Pre-Venture Entrepreneurs in Food Deserts

TL;DR

MISCON addresses the need for accessible, mission-driven guidance for pre-venture food entrepreneurs in food deserts by integrating a heterogeneous knowledge graph with multi-LLM orchestration within a state-space, milestone-driven conversation framework that encodes goals $G$, states $S$, milestones $M$, and transitions $T$. The approach supports adaptive intent management, milestone progression, and dynamic knowledge queries to tailor guidance across market, product, financing, and regulatory decisions. Key contributions include a formal state-space model, an architectural blueprint for a heterogeneous data stack with a virtual knowledge graph atop the AWESOME polystore, and a set of interaction primitives that balance factual accuracy, interpretability, responsiveness, and cost. This framework enables structured, multi-session consultations that persist user state, providing practical impact for the NOURISH initiative and a platform to study dialogue challenges in semi-open domains with large, heterogeneous knowledge bases.

Abstract

This work-in-progress report describes MISCON, a conversational consultant being developed for a public mission project called NOURISH. With MISCON, aspiring small business owners in a food-insecure region and their advisors in Community-based organizations would be able to get information, recommendation and analysis regarding setting up food businesses. MISCON conversations are modeled as state machine that uses a heterogeneous knowledge graph as well as several analytical tools and services including a variety of LLMs. In this short report, we present the functional architecture and some design considerations behind MISCON.

Figures (2)

• Figure 1: The simplified shows examples an ideation conversation with goals, subgoals and shifting intent. For lack of space, we do not elaborate on the subgoals.
• Figure 2: A sample intent hierarchy that is pre-learned by the system.