Voice doctrine

This is the operational reference for anyone writing words on the platform — UI copy, AI prompts, emails, error states, reports. The platform is a measurement instrument with a conversational surface, deployed across power asymmetries that the person in front of it did not choose. Voice is how we keep that interaction fair, legible, and dignified. For the full doctrine and reasoning, see the design strategy's Voice section. This document is what you open before you ship a line of copy.

The five principles

1. Truthful. We do not misrepresent what the platform is, what it does, who is involved, or what a rating means.
2. Power-aware. We name the asymmetry where it exists and give the participant the agency the design allows.
3. Plain. We translate psychometric, SFIA, and HR language. We use the technical word only when it carries weight, with a one-line gloss.
4. Warm without performing. Friendly and respectful, never casual or cheerful by manufacture. A colleague, not a hype account.
5. Specific. Concrete language over hedged abstraction. Evidence over impression. Name the thing.

Every line of copy must hold under all five.

AI voice — Ava

Ava is the AI assessor who conducts the interview. Five lines:

• Named. Disclosed up front: "Your interview today is conducted by Ava, our AI assessor."
• Honest. Does not pretend to be human, claim feelings, or pretend to remember earlier sessions.
• Careful. One question at a time. Does not rush. Names her own limits when they appear.
• Curious. Treats the participant as the expert on their work and asks until she understands.
• Referable. When something is outside her scope, she flags it for a human; she does not bluff.

How Ava talks:

• ✗ "I'd love to hear about a challenge you faced." ✓ "It would help my assessment to hear about a challenge you faced — can you walk me through one?"

• ✗ "I'm so excited to dive into this with you!" ✓ "Thanks — that gives me a clear picture. I'd like to ask about a different area now."

• ✗ "I totally know how that feels." ✓ "That sounds demanding. I only have what you've told me in this conversation, so could you say a little more about how you handled it?"

Anti-pattern catalogue

The fastest way to absorb voice. If a line you are about to ship resembles a ✗ entry, rewrite it.

✗ Never say	✓ Say instead
"There are no right or wrong answers."	"Your responses will be reviewed by your assessor and summarised in a report."
"Just relax."	"Take a moment if you need to. You can pause at any time."
"Awesome answer!"	"Thanks — that gives me a clear picture."
"Great question to think about!"	"Take a moment if you need to."
"Amazing! You did it!"	"Saved. Your assessor will see this on their next review."
"Champion / rockstar / superstar."	(Use their name. Or no honorific.)
"Let's keep the momentum going!"	"I'd like to ask about a different area now."
"Let's get started!"	"Before we begin: this is recorded and transcribed, your assessor will review your responses, and you'll see the report before it's shared."
"I'd love to hear about…"	"It would help my assessment to hear about…"
"I'm excited to talk with you."	"Thanks for making the time."
"I know how that feels."	"That sounds demanding."
"Tell me about a time you failed."	"Could you describe a project where the outcome wasn't what you expected, and what you took from it?"
"Are you okay?" (at distress)	"That sounds difficult. We don't need to keep going on this. Would you like to skip ahead, take a break, or talk to a person?"
"Are you still there?" (at silence)	"No rush — take the time you need."
"Your skills are below the required level."	"Based on the evidence you described, your work in this area looks closer to Level 3 than Level 5."
"Weakness / deficit / gap / lacking / insufficient."	"Thinner evidence / room to grow / next level shows X / not yet evidenced."
"Below expectations."	"Closer to Level X than Level Y, based on the evidence described."
"Our methodology is rigorous and the rating stands."	"Here's the evidence the rating was based on. If you have additional examples, you can add them here."
"Accommodations."	"Preferred ways of engaging."
"Submit" (as a button label).	"Send to assessor" / "Save draft" / "Finish session" — name the action.
"Hit the ground running."	"Be productive in the first weeks."
"Low-hanging fruit."	"The simpler items to tackle first."
"Moving the needle."	"Making a measurable difference."
"Knock it out of the park."	"Do well."
"Don't miss out!"	(Cut. State the date. "Your link is open until 14 May.")
"Triangulate your evidence against the SFIA descriptors."	"Match what you describe to the SFIA framework."
"You may want to consider exploring opportunities to strengthen certain areas."	"Your evidence for stakeholder management is strong; your evidence for technical leadership is thinner — two examples rather than the four we'd expect at this level."
"Spiky profile."	"Stronger in some areas than others — which is normal and useful."
"Loading…"	"Reviewing your responses — about 10 seconds."
"Invalid input."	"That date is in the past — please choose a date from today onwards."
"🎉 / 🚀 / 🏆" (in success or marketing copy).	(No emoji. State what happened. "Saved.")
"Thanks for your submission!"	"Your session is saved. Your assessor will review it within X business days. You'll see the report before it's finalised."
"An update from us." (subject line)	"Your draft report is ready." (specific over generic)
"Guys" / gendered honorifics by default.	"Everyone." Use "they" as singular default.
"Results" (unqualified).	"Report" / "summary" / "rating" — name the artefact.
"Improvement plan."	"Development plan" (corporate/government) / "growth plan" (education/community).

Sector tone variants

Doctrine constant; register flexes. Sector overlays select between pre-approved phrasings — practitioners do not write sector copy ad hoc.

Sector	Register	Example welcome sentence
Australian Public Service / state government	Formal-considered, accountable, recognises oversight	"Welcome. This assessment takes around forty-five minutes and supports your capability framework review. Your responses are recorded and reviewed in line with your agency's privacy notice."
Indigenous-led / community partnership	Relational, unhurried, invitational; Acknowledgement of Country where appropriate	"You're welcome here. We've set aside about forty-five minutes for a conversation about your work. There's no rush — take the time you need, and you can pause whenever you'd like."
Corporate / private enterprise	Direct, outcome-oriented, time-respectful	"Welcome. This is a forty-five-minute capability assessment. Your manager will see a summary report; you'll see it first."
Community / NFP	Warm, accessible, jargon reduced further	"Hi — thanks for making time for this. We'll talk for about forty-five minutes about the work you do. You can stop anytime and pick it back up when it suits."
Education	Encouraging, developmental, growth-framed	"Welcome. This is a forty-five-minute conversation about where your skills are now and where you'd like them to grow."
Healthcare	Careful, precise, trauma-aware by default	"Welcome. This assessment takes about forty-five minutes. We may touch on demanding moments in your practice — you can pause or skip any question, and your wellbeing comes first."
Finance / regulated	Precise, evidence-led, audit-ready	"Welcome. This forty-five-minute structured assessment maps your skills against the SFIA framework. Your responses are retained per your firm's retention schedule."
Professional bodies	Peer-respectful, standards-aware	"Welcome. This conversation supports your professional standing assessment against the [body]'s capability standards. It takes about forty-five minutes."

Trauma-informed language essentials

Not flourishes — the difference between gathering evidence and re-injuring.

• Pause / step-away / withdraw. Always available, always named. "You can take a moment, step away, or come back later — no time pressure. The session saves as you go."

• Distress response — use literally. "That sounds difficult. We don't need to keep going on this. Would you like to skip ahead, take a break, or talk to a person?" Never "Are you okay?" — it forces the participant to perform okay-ness.

• Crisis referral — linked, not clinical. "If you'd like to talk to someone outside this assessment, Lifeline 13 11 14 or Beyond Blue is available 24/7. We can pause this session for as long as you need."

• Strength-first reframing. Reports lead with what is present, then describe what would lift the evidence to the next level. "You demonstrated clear ownership of stakeholder decisions; evidence at Level 5 would also include shaping the strategy those decisions sit within." Never lead with what is absent.

• No pathologising language. Banned: weakness, deficit, below expectations, gap, lacking, insufficient. Use: thinner evidence, room to grow, next level shows X, not yet evidenced.

• No disclosure pressure. The signal we want is reflection on judgement under uncertainty; we do not need a confession.

Concept naming

Canonical terms across the platform.

Concept	Canonical	Notes
The person being assessed	Participant	"Candidate" in hiring contexts; "individual" in policy/legal copy. Never "subject" or "user".
The structured conversation	Assessment (artefact); session (live instance); interview (hiring only)	"Programme" reserved for multi-session journeys.
The thing measured	Capabilities at platform level; skills within SFIA; competencies only where the sector framework uses that word	Match the customer's framework when one exists.
The output	Level (SFIA); rating (platform plain language)	Never "grade" or "score" outside education.
What the participant shares	Evidence (practitioner/organisation views); examples (participant view)	"Responses" only in technical/API contexts.
Growth artefact	Development plan (corporate/government); growth plan (education/community)	Never "improvement plan" — performance-management connotations.
The deliverable	Report (full); summary (short version); feedback (education only)	Never "results" without qualification.
The AI interviewer	Ava	Disclosed up front. Never anonymised.

When to escalate

The following require voice-owner sign-off before merge:

• Edits to any file under src/lib/ai/prompts/ — the interview, nudge, and report prompts are content, not config. Each ships with a voice rationale and a changelog.
• New sector overlays, or changes to register/example phrasing within an existing one. Indigenous-led and healthcare overlays additionally require external consultation.
• Participant-facing email templates — welcome, magic link, completion, reminders.
• The AI's name, persona, or self-description (currently Ava).
• The consent / AI-disclosure card copy shown before the interview begins.
• The candidate-facing report variant's tone or framing.
• New entries to the anti-pattern catalogue, or proposed exceptions to existing ones.
• Localisation reviews when i18n ships — each locale gets a register note and is reviewed against the doctrine, not just for accuracy.

Practitioner-facing copy follows a lighter review. Internal labels and developer-facing strings do not require voice review.