ENTERPRISE PROFILE STARTER KIT
Blank working templates for a human-led discovery engagement. No upload or processing service is activated by downloading or completing these files.

1. Agree one business outcome, workflow boundary, sponsor, accountable reviewer and information-handling scope first. Keep completed files in your organisation's approved workspace. Do not put confidential information into unapproved tools or email.
2. Use stable, unique, organisation-assigned IDs. Suggested prefixes: CAP-, PROC-, ROLE-, APP-, DATA-, POL-, KPI-, EVID-. These are conventions, not enforced software rules.
3. Every owner_role_id and approval_role_ids reference must resolve to a role_id in enterprise-roles.csv. parent_capability_id resolves to capability_id. Application/data/policy/capability/process references resolve to their respective files. applies_to_ids may refer to any declared entity ID.
4. Separate multiple IDs within a cell with a semicolon. source_refs contains evidence_id values from enterprise-evidence.csv. integration_refs references documented interface IDs in approved evidence; this starter kit does not include an interface schema.
5. An evidence approved_reference identifies an approved internal document or record location, not an access grant. Record version, date, owner, classification and permitted purpose. Never include passwords, API keys, access tokens or signed URLs. Do not copy sensitive documents into these templates.
6. review_status: unreviewed, proposed, accepted, disputed or unknown. These are declarations for review, not independently validated facts. Leave unknown values empty; do not infer approval or invent baselines.
7. measurement_status: measured, estimated or unknown. Values need a unit, period, definition and supporting source. Targets are aspirations, not measured benefits. differentiation can describe core advantage, supporting capability or an unresolved judgement.
8. Minimise personal information: use role identifiers rather than employee names or personal addresses where possible. Apply your organisation's retention, access, privacy and legal rules.
9. CSV files contain headers only. Use UTF-8 and standard CSV quoting for commas/newlines. Treat spreadsheet formula prefixes (=, +, -, @) in imported text as potentially executable; import untrusted cells as plain text. This kit has no automated import or sanitisation.
10. Reconcile duplicated IDs, dangling references, conflicting policies, source freshness, ownership and missing evidence before accepting a contextual model. A named human reviewer records the decision. An AI suggestion never grants access, resolves authority or authorises an enterprise action.

PROPOSED FUTURE STUDIO
Authenticated tenant workspaces, approved-source connectors, uploads, deterministic validation, model-assisted extraction, review and exports are roadmap concepts, not operational features in this release. Their implementation requires agreed scope, security/privacy design, provenance, evaluation, resource limits, recovery, approval and verification. Service availability, deliverables, duration and pricing are to be agreed.

Next step: discuss the scope at https://contextualagentic.ai/contact/