Reference values, not opinions
Quantity decisions use documented nutrient reference values and approved upper limits. Where a limit applies, it is enforced deterministically rather than left to interpretation.
Science and evidence
Personalisation is only trustworthy if it is governed. Our recommendations are produced by approved rules, documented sources and reproducible logic — not by an unconstrained model.
Authorised claims only
Customer-facing wording is drawn from a controlled list of approved claims. Prohibited claims are recorded alongside each goal module so they cannot be reintroduced.
Versioned rules and sources
Safety rules, interaction rules, goal definitions and source documents are versioned with effective dates, review dates and a named professional approver.
Full provenance
Every generated plan stores a snapshot of the exact questions, answers and rule versions used, so any recommendation can be reconstructed and audited later.
Evidence hierarchy
We distinguish a customer goal, a possible nutritional consideration, a product candidate, a permitted recommendation and a professionally reviewed plan. These are never blurred together.
Honest limitations
Nutrition evidence is incomplete and individual responses vary. Where the evidence does not support a recommendation, we do not make one.
Where optional AI guidance fits
Conversational guidance can help you understand your plan. It never sets quantities.
- AI guidance explains approved content and helps you navigate the assessment.
- It cannot create, alter or override a recommendation, quantity or safety rule.
- It cannot diagnose, interpret laboratory results or advise on medicines.
- Prompts and model configurations are versioned and published only after approval.
- Where a question needs a professional, the conversation is escalated rather than answered.