Reported enterprise deployment readiness of privacy-preserving artificial intelligence: a governance and information management perspective
Ashrafe Alam, Md Golam Rabbani, Gahangir Hossain, Victor PrybutokPurpose
Privacy-preserving artificial intelligence (AI) technologies increasingly support data-driven decision-making while safeguarding sensitive information. However, peer-reviewed evidence documenting governance integration and operational deployment remains limited. This study examines how the peer-reviewed academic literature documents the technical maturity, governance integration, and operational implementation of privacy-preserving AI systems relevant to enterprise information systems.
Design/methodology/approach
A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020-compliant systematic literature review examined 22 peer-reviewed studies published from 2018 to May 2026. The review synthesized evidence on privacy–utility trade-offs, governance integration, fairness, operational readiness, and implementation maturity. We used Technology Readiness Levels (TRLs) and a proposed Translational Readiness Index (TRI) to classify readiness evidence reported in the included studies.
Findings
Privacy-preserving AI systems demonstrated favorable privacy–utility trade-offs in the reviewed academic literature, with privacy budgets of e = 1–3 preserving approximately 85–96% of baseline predictive performance. Federated learning combined with differential privacy was the dominant implementation architecture. Most included studies demonstrated intermediate technical maturity (TRL 4–6), primarily through controlled evaluations and prototype development while none documented deployment within operational payer workflows. Studies also inconsistently documented governance integration, privacy accounting, interoperability, and fairness assessment. We characterize this imbalance in the published academic evidence as the proposed Enterprise AI Documentation Gap Hypothesis. The hypothesis concerns the distribution of evidence in peer-reviewed literature and should not be interpreted as evidence that healthcare organizations or insurers are unprepared for deployment.
Research limitations/implications
This study is limited to reviewing scholarly papers published in the academic literature and did not systematically examine industry literature, company records, patent information, or practitioner views. Therefore, the results reflect the scholarly evidence base and do not reflect health organizations' or insurers' readiness for implementation. The absence of evidence in the literature on deployment does not imply that implementation does not occur outside scholarly journals. Future studies should cross-check scholarly evidence with industry and practitioner evidence on the same topic.
Originality/value
The TRI offers an exploratory, multidimensional framework for organizing documented evidence across technology, privacy, governance, fairness, and operations. This study proposes the Enterprise AI Documentation Gap Hypothesis to describe the gap between the relative maturity of technical evidence and the lack of published evidence on governance and operations. These are conceptual contributions that require further empirical testing.