Classical Intelligence For Quantum Technologies: An Evidence Roadmap For AI‐Enabled Quantum‐Computing Workflows

This is the authors' abstract. We don't add key points for this paper.

Artificial intelligence (AI) is becoming an operational part of quantum‐computing workflows. Here, classical intelligence denotes classical learning, inference, and adaptive search—machine learning (ML), deep learning (DL), and bio‐inspired optimization—that improves how a quantum workflow is controlled, executed, or interpreted. A structured narrative review maps these methods across the control, readout, compilation, and application layers. A reproducible search identified 379 records; 112 directly evaluative studies and 21 contextual records were retained. Each study was coded on four reporting dimensions as explicit, partial, or absent. Visibility was asymmetric (count, 95% Wilson interval): quantum budget 109/112 (92.4%–99.1%), classical overhead 72/112 (55.1%–72.6%), robustness 70/112 (53.3%–70.9%), and transfer 56/112 (40.9%–59.1%). A separate deployment‐readiness judgment, derived from the evaluation regime, places 85/112 studies at simulator‐only, 13/112 at mixed or cloud, and 14/112 at hardware validation; none met the controller‐service criterion. A linked results layer records each study's comparator and direction of effect. ML is most defensible when uncertainty or reusable structure reduces quantum queries, DL when raw or high‐dimensional structure dominates, and bio‐inspired search when objectives are rugged or mixed discrete‐continuous. The field's main limitation is incomplete end‐to‐end evidence on classical cost, drift, transfer, and deployment, not an absence of benefit.