The line above captures the problem federal organizations confront as they push artificial intelligence into mission-critical work: getting data is necessary, but it is not sufficient. In a recent episode of the Government Technology Insider podcast, host Lucas Hunsicker and Kevin Bohan laid out a specific, practical gap—semantic trust—that sits between governed access and reliable AI-driven decisions.
How inconsistent definitions break AI reasoning
Bohan explained that even when an AI system obtains the data it requires, it may still “not always understand it accurately.” The root cause, as he described it, is not missing access or blocked permissions but the absence of a shared vocabulary: different systems use different definitions, terminology, and metrics for what should be the same concept when describing the same people, assets, programs, and missions. The result is inconsistent or contradictory outputs from AI systems that are otherwise fed correct data.
Why semantic trust is distinct from access and governance
The conversation drew a clear distinction between three layers of trust. Access and governance determine whether data can be seen and who can use it; semantic trust ensures that what is seen is interpreted consistently. Bohan argued that without that shared meaning, governed data can still produce divergent results across departments and mission partners, undermining coordinated decision-making.
The cyber insurance questionnaire just landed. Now what?
SOC 2, HIPAA, insurance renewals - someone has to own security strategy. Nubivance provides fractional CISO leadership without the full-time salary.
Get a security leadImplications for agencies embedding AI into mission-critical operations
The podcast framed the issue in operational terms: federal agencies are embedding AI into mission-critical operations, and the stakes are organizational consistency and reliable automated reasoning. According to the discussion, agencies investing heavily in AI should treat access and governance as necessary steps but not the end goal; achieving trustworthy AI requires establishing semantic trust as the foundation for consistent decision-making across the enterprise.
Lucas Hunsicker, Kevin Bohan, and Denodo’s framing
The exchange on Government Technology Insider—hosted by Lucas Hunsicker and featuring Kevin Bohan, Director of Product Marketing at Denodo—focused on the practical mechanics of that framing. Bohan described how misinterpretation occurs when systems lack shared definitions and how those misinterpretations fuel disconnects between departments and mission partners, highlighting semantic alignment as the missing layer between governed access and dependable AI outputs.
What this means for federal agencies, departments and mission partners, and AI governance teams
- Federal agencies: Agencies investing in AI should prioritize creating and enforcing shared definitions and metrics so AI systems can reason consistently across programs rather than rely solely on governed access.
- Departments and mission partners: When partners use different terminology or metrics for the same entities, AI-driven coordination can produce contradictory results; aligning on meaning reduces those operational disconnects.
- AI product and governance teams: Teams responsible for data governance and AI productization must treat semantic trust—consistent understanding and current information—alongside traditional governance controls over who can access and use data.
Building semantic trust is a concrete, technical complement to existing governance and access controls: it is the shared meaning that lets multiple systems and stakeholders interpret the same records in the same way. If agencies heed the argument laid out by Bohan on Government Technology Insider, the next phase of AI adoption will emphasize semantics—definitions, terminology, and metrics—so that correct data yields correct conclusions, not merely consistent access.
Read the original Government Technology Insider podcast summary
