Data governance gaps: what they mean for enterprise AI teams

TL;DR: Enterprise AI stalls less because of model quality than because consent, preference, and permission data are fragmented, enforced inconsistently, and too slow to support real-time AI pipelines, according to Transcend; it cites MIT's 2025 GenAI Divide study, covered by Fortune, which found 95 percent of enterprise generative AI pilots failed to deliver measurable financial return. Governance has to move from paper policy to system-level enforcement if AI is going to scale safely.

Editorial analysis by NHI Mgmt Group, based on content published by Transcend: “Is your data governance strategy slowing your AI roadmap?”.

By the numbers:

• MIT's 2025 GenAI Divide study, covered by Fortune, found that 95 percent of enterprise generative AI pilots failed to deliver measurable financial return.

Key questions

Q: How do security and privacy teams stop AI from using data that is not approved?

A: They need to enforce consent, purpose, and retention rules at the point data is accessed, not after a model has already consumed it.

Q: Why do fragmented permissions create risk in enterprise AI?

A: Because AI decisions depend on knowing whether a dataset is actually usable for a specific purpose at a specific moment.

Q: What breaks when consent changes do not propagate to downstream systems?

A: If consent updates do not reach CRM, email, analytics, and customer data platforms quickly, an organisation can continue processing after a user has opted out.

Practitioner guidance

• Implement real-time permission enforcement Apply consent, opt-out, and purpose checks before data is handed to model training, analytics, or marketing systems so downstream use reflects current policy.
• Unify data discovery and inventory Maintain a continuously updated catalogue of where personal and governed data lives, how it moves, and which systems can process it.
• Propagate preference changes automatically Ensure Do Not Train, deletion, and communication preferences update across warehouses, SaaS tools, and AI pipelines without manual rework.

Bottom line: Enterprise AI often stalls because governance is documented but not technically enforced across the systems that move and reuse data.

Explore further

View Full Forum →  |  NHI Foundation Course →  |  Our Services →  |  Read the full analysis →