A confident sentence can still be wrong.

How SI decides what it can responsibly publish.

SI publishes canonical doctrine, evidence-led analysis, research summaries, executive interpretation, Founder perspectives, client evidence stories, illustrative demonstrations and promotional content. These do not carry the same evidential weight.

Content classes
Canonical doctrine

Approved SI architecture and definitions.

Evidence-led analysis

Interpretation grounded in identified evidence and visible limitations.

Research summary

Synthesis of attributed external sources.

Executive interpretation

Professional judgement presented as interpretation.

Founder perspective

A named viewpoint that does not automatically become canonical doctrine.

Client evidence story

Permissioned or appropriately anonymised, with retained evidence.

Illustrative method demonstration

Hypothetical or representative, labelled as not a client result.

Promotional content

Still subject to evidence and claim controls.

The claim test

For every material claim, we record the source, period, scope, method, evidence status, limitation, reviewer, approver and review date.

Evidence statuses
Observed Inferred Modelled Validated Unknown

Unknown must not be removed merely because it makes the story less convenient.

AI-assisted content

AI may organise, compare, summarise, draft, edit and support preliminary analysis. It may not approve its own output, invent evidence, infer permission, alter canonical doctrine, expose confidential information or convert Inferred into Validated.

Corrections

When a correction is required, we record the previous wording, corrected wording, reason, date, owner, affected assets and supersession.

Say what the evidence supports. Then make the boundary equally visible.