CAi Copilot: Reducing Operational Workload in Molecular Design through Intent-Driven Agentic Workflows

2026-08-07

AI Summary

Early molecular design relied on manual coordination of disparate AI tools, resulting in fragmented workflows and substantial cognitive burden. This work proposes the first intent-driven, three-layered agent architecture—comprising a research interface layer, a reasoning layer, and an execution substrate—that automatically translates ambiguous research objectives into executable, traceable, and dynamically adaptable molecular design workflows. The framework seamlessly integrates modules for molecular generation, property prediction, molecular docking, and synthetic feasibility assessment. Evaluated across 45 tasks, the system achieves a composite score of 84.59, significantly outperforming the next-best approach by 18.07 points, thereby demonstrating its efficacy and interpretability in multi-criteria collaborative optimization.

Abstract

Early-stage molecular design is an iterative process, not just a task of generating molecules. Researchers turn broad goals into design strategies, refine candidates, assess many properties, and gather evidence before synthesis and tests. AI methods can generate molecules, optimize several goals, predict properties, dock compounds, and account for synthesis. Yet these functions are spread across specialized tools. Experts must still coordinate each step, judge interim results, and integrate evidence. The central challenge is thus to turn research intent into adaptive, traceable runs grounded in scientific tools. We cast this challenge as intent-to-evidence molecular design workflow execution and present CAi Copilot, an expert-oriented agent with three linked layers. The Research Interface Layer turns intent into an executable plan. The Agent Reasoning Layer uses interim results to guide each run. The Execution Substrate supplies molecular tools, metrics, reusable utilities, and backend services. Across 45 tasks, CAi achieves the strongest overall performance, with an outcome score of 84.59, exceeding the next-best result by 18.07 points. Additional benchmarks test how CAi coordinates generation, screening, and multi-criteria evaluation, while exposing limits in long-horizon execution. These results show that CAi turns broad molecular-design intent into transparent, traceable workflows that connect interim decisions to candidate-level evidence.

Problem

• molecular design
• intent-driven workflow
• operational workload
• evidence integration
• agentic systems

Innovation

• intent-driven workflows
• agentic molecular design
• adaptive execution
• traceable reasoning
• expert-oriented AI