Composite AI With Custom, Compact, Neurosymbolic Models: The Emergent Enterprise Artificial Intelligence Paradigm

Amit P. Sheth, Kaushik Roy, Revathy Venkataramanan, Venkatesan Nadimuthu, Chathurangi Shyalika

Abstract: Artificial-intelligence architectures are shifting from monolithic, internet-scale models toward multi-component composite systems that must perform reliably in high-stakes settings. We introduce Custom, Compact and Composite AI with a Neurosymbolic approach (C3AN), a fourth-generation framework that integrates data, domain knowledge and human expertise through 14 foundation elements spanning reliability, grounding and safety. Custom focuses on right-sized, domain-specific data and workflows; Compact emphasizes resource-conscious models that can run on edge or mobile hardware; Composite unifies neural, symbolic and feedback modules for end-to-end reasoning. Three pilot systems illustrate the paradigm: Nourich (diabetes-aware dietary guidance) outperforms nine LLM baselines on recipe suitability; MAIC (K–12 mental-health triage) matches PHQ-9 gold-standard accuracy while halving compute; and SmartPilot (edge manufacturing copilot) achieves 93% anomaly accuracy and 21% forecasting gain over LSTM baselines. C3AN demonstrates that domain-aligned, neurosymbolic design can deliver transparent, trustworthy enterprise AI without extreme scale.