1
Input
A document exists, but the readable text is incomplete or unreliable.
Tenant Packs
AI becomes useful when it knows what it knows, what it does not know, and in which context it is operating.
Tenant Packs provide the starting context for a specific knowledge environment. They bring foundational knowledge, scope, and governance so AI does not begin from zero.
A familiar failure
When context is weak, AI still sounds certain.
A scanned agreement arrives with a broken text layer. A generic assistant may still summarise, translate, or answer questions as if the document were fully readable. That is exactly where trust breaks down.
A reliable system should first make readability limits explicit.
2
Typical response
The model fills gaps with plausible language and presents it as fact.
3
What should happen
The system should mark uncertainty, preserve provenance, and ask for a better source when needed.
What changes
Tenant Packs turn a generic model into a context-aware starting point.
A Tenant Pack does not claim truth. It defines what kind of knowledge environment the system is in before answers are generated or decisions are supported.
Without a Tenant Pack
The model starts from broad world knowledge, loose assumptions, and little understanding of local rules.
With a Tenant Pack
The system starts with scoped terminology, relevant relationships, access logic, and the right level of caution for that domain.
What a pack contains
A Tenant Pack is a configurable context and governance package.
It does not replace the model. It shapes the starting conditions under which the model can become useful.
Foundational knowledge
Key concepts, basic relations, and what counts as normal in a domain.
Context boundaries
What belongs to the knowledge space and what lies outside it.
Governance rules
Who may see, review, or elevate information into trusted use.
Domain language
Relevant terms, meanings, and distinctions for that environment.
Evidence patterns
What kind of support is expected before a claim should be relied on.
Responsible defaults
The right level of caution for ambiguity, missing data, or weak sources.
Where it matters
The same principle can be reused across very different knowledge environments.
The architecture stays the same. Only the starting context changes.
Enterprise operations
Contracts, handovers, project history, internal decisions, and team memory.
Sensitive personal context
Longevity notes, routines, private records, and other highly personal knowledge environments.
High-trust collaboration
Compliance-relevant reviews, research workflows, and contexts where provenance matters before action.
Why this matters
Tenant Packs make governed AI usable without pretending every context is the same.
They are the starting layer that helps an AI system say what it knows, what remains open, and what must be reviewed before trust is justified.
Reusable across domainsSensitive-context readyContext before output