Topic starter
02/08/2026 11:34 am
TL;DR: Data debt, incomplete classification, unscanned storage, and ungoverned access are the pattern behind many AI project misses, according to Mind. AI systems inherit the condition of the data estate, so weak governance turns into weak outputs and missed KPIs.
Editorial analysis by NHI Mgmt Group, based on content published by Mind: “Why most AI projects are failing”.
Key questions
Q: Why do enterprise AI programmes fail even when the model performs well?
A: Because model accuracy does not solve access, workflow, data, and accountability problems.
Q: What are the signs that AI governance is failing in the enterprise?
A: Common warning signs include rapid growth in AI use without matching policy coverage, sensitive files being copied into personal accounts, and a large share of AI apps carrying high or critical risk.
Q: How should organisations keep AI pipelines from inheriting data governance gaps?
A: Treat the AI data path as part of the governance boundary.
Practitioner guidance
- Define AI data trust prerequisites Set minimum standards for classification, provenance, and ownership before any AI use case is allowed to consume the data source.
- Review inherited permissions for AI-connected data Audit the access model on repositories connected to AI tools, including permissions inherited from legacy cleanup or incomplete offboarding.
- Measure trust, not just activity Pair prompt volume, token counts, and usage telemetry with metrics for data quality, catalog coverage, and access scope.
Bottom line: Most AI projects are failing because weak data governance, not model quality, is breaking trust in the outputs.
Glossary terms: Data Trust Contract, Inherited Access, Non-Human Identity (NHI), Identity Trust Debt
Explore further
View Full Forum → | NHI Foundation Course → | Our Services → | Read the full analysis →
This topic was modified 2 days ago by NHI Mgmt Group
02/08/2026 12:48 pm
Data trust is becoming the gating control for AI governance: the primary reason AI programmes miss business outcomes is not model sophistication but weak control over the data feeding the model. Classification gaps, stale repositories, and unmanaged access create a reliability problem before AI ever becomes a security problem. The organizations that treat data trust as a prerequisite, not a post-launch cleanup task, are the ones more likely to produce outputs the business can actually use.
A few things that frame the scale:
- Companies using AI governance put over 12 times more AI projects into production, according to a Databricks report.
A question worth separating out:
Q: Should AI data access be reviewed differently from normal application access?
A: Yes. AI-connected access should be reviewed as a distinct control problem because the system can reach broad data estates, inherit permissions from old cleanup work, and expose information at machine speed. The question is not only who can log in, but what the model can legitimately see.
👉 Read our full editorial: AI projects fail when data trust and access governance lag