Governments can run many AI pilots and still fail to become more intelligent.
The value of a government experiment does not lie only in launching a model or service. It lies in whether the institution records its assumptions, measures outcomes, examines failure, validates conclusions and carries what it learned into later policy, procurement and service decisions.
Evidence
The OECD analysed 200 government AI use cases and identified practical constraints including skill gaps, legacy systems, limited data, tight budgets and stronger public-sector requirements for privacy, transparency and representation.
The GovTech Maturity Index reviewed public-sector digital transformation across 197 economies. Although it collected information on the use and performance of government systems and platforms, only a few economies provided strong evidence—indicating weak monitoring and reporting.
The government AI playbook requires meaningful human control, full life-cycle management, robust assurance, pre-deployment testing and regular checks after deployment.
NIST's Generative AI Profile places governance, pre-deployment testing and incident disclosure among its primary considerations, and calls for empirical evaluation and documented testing across the AI life cycle.
Interpretation
These sources do not establish the Sixth Generation of Digital Transformation. Taken together, I interpret them as evidence of a deeper managerial problem: an AI project may produce outputs without creating durable institutional learning.
Proposed 6GDT application
Within my proposed Sixth Generation of Digital Transformation (6GDT) framework, adoption is not enough. Government must build an applied learning loop:
This is a domain-specific application of 6GDT, offered for research and testing rather than as an established government-maturity model.
Generation is not inference
Generative Intelligence can draft text, summarise records and propose possible responses. In the proposed 6GDT framework, Inferential Intelligence goes further by helping formulate hypotheses, identify hidden relationships, examine possible causal explanations and test alternative interpretations through Human–AI Co-Discovery.
It does not replace public accountability or managerial judgment. Inference must remain open to challenge, evidence, validation and human responsibility.
The managerial test
Before calling an AI initiative successful, a public institution should be able to answer five questions:
- What public problem and baseline did the pilot define?
- Which assumptions, evidence and errors were recorded?
- What validation threshold determined whether the result was credible?
- Who owns the negative findings and transfers them across agencies?
- How will the validated lesson change the next rule, budget, procurement or service?
Public value
When learning becomes institutional memory, AI can support more than efficiency. It can reduce repeated expenditure, preserve decision rationale, strengthen cross-agency learning and improve the evidence available for allocating public resources.
Foresight proposition
The future divide may not be between governments that possess AI and those that do not. It may be between governments that repeatedly deploy technology and governments that continuously learn from it. This is a testable proposition—not an empirical conclusion.
Neo-Inference Science (NIS) is proposed as the theoretical foundation of this framework. It remains a developing research program that requires independent conceptual and empirical validation.
Sources
- OECD — Governing with Artificial Intelligence (2025)
- World Bank — GovTech Maturity Index 2025
- UK Government — Artificial Intelligence Playbook for the UK Government (2025)
- NIST — Generative Artificial Intelligence Profile, NIST AI 600-1 (2024)
Evidence, interpretation and the proposed framework are separated deliberately. The cited institutions do not endorse 6GDT or NIS.
