What does your evidence actually support?
Give Loosehead a defined question and the documents or data available to answer it. Loosehead determines what the supplied evidence supports, what it does not, and where proceeding would require missing evidence, meaning, authority or an unsupported assumption.
Deterministic evaluation. No invented evidence. No automated business decision.
A question. The available evidence. A bounded determination.
Organisations already hold large amounts of information in documents and data. The difficult question is often whether that information actually supports the conclusion someone — or some system — is being asked to reach.
What is supported
Loosehead identifies where the supplied evidence satisfies the conditions required by the defined question.
What is not supported
Missing or inadequate evidence is surfaced rather than silently repaired, enriched or converted into an assumption.
What is still required
Where the question depends on missing evidence, meaning or authority, Loosehead makes that boundary explicit.
Question + evidence → Loosehead → what the evidence supports.
Loosehead evaluates a defined question against the evidence that was actually supplied. It preserves the distinction between what is present, what is absent and what cannot legitimately be inferred.
Question + evidence
A bounded question together with the information currently available to answer it.
- Documents and policy material
- Structured data and records
- Schemas, metadata and mappings
- Declared conditions or requirements
Loosehead
Required conditions are evaluated against the supplied evidence. The core determination is rules-based and deterministic; missing meaning or authority is not invented in order to force an answer.
Attributable result
The result records what can and cannot be established from the evidence available.
- Supported
- Not supported
- Cannot be established
- Basis preserved for review
The evaluator does less — deliberately.
Many systems score, enrich, repair, infer, optimise or recommend. Loosehead preserves a narrower boundary: determine what supplied evidence supports without converting that determination into authority or action.
Deterministic core
The same bounded evidence, declared conditions and rules are intended to produce the same evaluative result. Probabilistic AI is not required for the core determination.
Non-inventing
Missing evidence, meaning or authority is surfaced. It is not silently filled in simply to allow evaluation to continue.
Non-authoritative
Loosehead determines what the evidence supports. The organisation remains responsible for policy, authority, interpretation and consequence.
Evidence before conclusion.
Determination before decision.
Wherever an answer depends on evidence.
The subject can change while the underlying problem remains the same: a defined conclusion is being asked of documents, data or both, and the organisation needs to know whether the available evidence actually supports it.
Policy & contractual material
Evaluate a defined question against terms, policies, standards, requirements or other controlled documents.
Operational decisions
Determine whether the records available at a decision point support the stated conclusion or required conditions.
Data & model inputs
Establish whether supplied data carries the evidence and meaning a proposed downstream use expects.
Transformation & integration
Determine whether required meaning survived mapping, translation, system boundaries or representation change.
Automation & AI
Establish the evidential basis available before automated reasoning, recommendation or consequential action begins.
Audit & review
Preserve an attributable record of what could and could not be established from the evidence available at the time.
See Loosehead work.
We are preparing a browser-based demonstration of Loosehead using bounded synthetic cases.
Start with a defined question
The demonstration begins with a specific conclusion or use to be tested rather than an open-ended request for an answer.
Supply bounded evidence
Documents, structured information and declared conditions are kept within an explicit evidence boundary.
Inspect the determination
See what is supported, what is not, what is missing and the basis for the result.
See where AI stops
AI can assist at defined boundaries. It does not get to invent the evidence, meaning or authority required for the deterministic determination.
Bring one real question.
A first engagement can be deliberately small: one defined question and the documents or data your organisation currently uses to answer it.
A null result is valid. The purpose is to establish what the supplied evidence supports, not to force a predetermined answer.
You bring the question
Choose a bounded problem where the answer depends on evidence your organisation already holds.
You supply the permitted evidence
Documents, data or both. The boundary is agreed before evaluation.
Loosehead evaluates it
The evaluation remains bounded to what was supplied and to the declared conditions relevant to the question.
You receive the basis
A transparent record of what is supported, not supported or cannot be established — and why.
Built through governed experimentation.
Loosehead has been developed by separating product meaning from implementation choice, testing the same underlying evidential boundary across different technical and operational settings, and preserving evidence from each pass.
Have a question whose answer depends on evidence?
Bring us one defined question and the documents or data you currently use to answer it. We are looking for organisations willing to test Loosehead against tightly bounded real-world problems.