In-App Microcopy — consent checkpoints, dialogs, guidance

(The plain-language User Guide remains the binding tone/terminology spec; these are the mandatory legal and safety surfaces.)

Signup & consent

Tenant creation (blocking):

Your data, in order. Before we begin: the Data Processing Agreement makes your organisation the controller of everything you assess and us your processor — your names and amounts never leave the system, everything resides in the EU, and every number has an audit trail. [Review and accept the DPA] (required)

Telemetry consent (tenant admin, optional, default off):

Help calibrate the methodology? Share anonymised, aggregate-only statistics (score distributions, warning frequencies — never names, amounts or documents) to improve the published calibration. Withdraw any time; nothing about your group is identifiable. [Yes, contribute] [No]

Evidence upload

Attach the best single document — one strong contemporaneous record beats five memos. Before uploading: remove personal data the evidence doesn't need (redact salaries, private details); never upload health, criminal-record or other special-category data. Files are sealed in your tenant's vault — even our operators cannot read them without a logged, justified elevation you can audit.

Contributor invitation (e-mail + first screen)

You've been asked to answer [n] questions about [area] — because you're the person who actually knows. You'll see your questions and nothing else: no scores, no conclusions, no other entities. Your answers enter the audit trail under your name, which makes them evidence, not hearsay. Due: [date].

Question assignment reminder

[Question] is waiting for you (due [date]). Two minutes of your knowledge now saves the group an inspector's question later.

Warning pattern (every coherence/creep/gate message: FINDING + ACTION)

Template:

Finding: [what the engine detected, plain language, with the number]. Why it matters: [one sentence, the rule's rationale]. What to do: [the concrete next step: fix an input error / attach evidence / record a justification / escalate to review]. Example (mass balance): Finding: Claimed execution on "Core algorithm development" sums to 132% across entities — more work than exists. Why it matters: two entities cannot each perform most of the same activity; an inspector will ask which claim is wrong. What to do: revisit the execution scores for DevCo and ParentCo; if work genuinely shifted mid-year, use period-split scoring instead.

Lock confirmation (after D2 disclaimer)

Locked. [Year] for [Group] is now a permanent snapshot (hash [short-hash], methodology [vX.Y]). Next year starts from here — and only asks what changed.

Export dialog

Your data, portable. This export contains everything of yours: structured assessment data, scores, evidence metadata, computed results, and locked reports (machine- and human-readable). It does not include the methodology's proprietary parameters (question bank, rubrics, weights) — those stay ours so they can stay defensible for everyone. [Export tenant data]

Version migration prompt (never automatic)

Methodology [vX.Y] is available (you're on [vA.B]). Migrate this year's assessment when you're ready — we'll show exactly what changed and split your year-over-year delta into what your group changed and what the methodology changed. Locked years are never touched. [See what's new] [Migrate this assessment] [Stay on vA.B]

Advisor access grant (client-initiated)

Grant [Firm] access to your tenant as [role], until [date/revoked]. They see what the role sees; every access is logged; you can revoke in one click. [Grant access]