Responsible technology use

AI may assist the work. It does not decide the person.

Our planned AI features are designed for drafting, structure and preparation. They are not hiring-decision tools.

Last reviewed: August 2026

This notice explains the intended safeguards for AI-supported features. It should be read with our privacy notice and candidate data page. The EU’s guidance on AI Act transparency obligations informs our commitment to clear notice where people interact with or receive content from relevant AI systems.

What AI may be used for

  • Drafting concise CV, LinkedIn, application and interview wording from candidate-approved facts.
  • Structuring a Career Storyboard entry or Evidence Bank record.
  • Generating practice questions and coaching prompts for an Interview Studio session.
  • Suggesting a field-mapping preview when a candidate imports an approved template.
  • Helping a candidate identify questions, gaps or next actions for human review.

AI-generated suggestions will be presented as suggestions. Candidates and consultants retain responsibility for reviewing facts, context, tone and suitability before use.

Arya: optional employer discovery

Arya is an employer research workspace inside the authenticated candidate and employee portals, separate from the public Cara assistant. When no AI provider is configured, guided mode opens suggested search links using your target role and chosen location. You choose an organisation and supply its public contact page. Arya checks that page for contact details and asks you to review them before saving. This mode does not send your documents to an AI provider. Opening an external search link shares its search terms with the search provider.

When automatic AI discovery is configured and you explicitly confirm, Arya uses relevant profile text, active Job Matrix targets, non-private Storyboard entries, Evidence Bank content and readable text from up to four recent supporting documents to research a location. Direct email addresses, long numbers and links in that career context are removed before it is sent, but free-text documents can still contain personal information. Do not upload sensitive personal information for research.

Eligible configured AI research routes include Perplexity and OpenAI. Automatic mode can try another configured provider if the first fails. Gemini native Google Search grounding is not enabled for this database workflow. Arya checks public source pages before saving supported contact routes to the candidate’s private Employer Database. AI-generated relevance suggestions are not proof of vacancies, hiring intent or guaranteed outcomes. Missing or unsupported contact details remain blank. Confirm details before using them.

Employees can use Arya only for candidates they are actively allocated to support; administrators and owners have wider operational access. Staff access, document uploads, research requests and outreach updates are recorded. Saved records remain when you start a fresh search. Arya never sends outreach automatically. The optional daily goal counts organisations explicitly marked as contacted, not research results, drafted emails or applications. You can avoid external AI processing by using guided mode where offered, or by not starting an AI search.

What AI must never do

  • Assess or infer protected characteristics, including health, disability, ethnicity, religion, political views, sexuality, age or other sensitive traits.
  • Score employability, decide whether someone is suitable for work, or create a personality or risk score.
  • Rank, reject or recommend candidates to employers as an automated hiring decision.
  • Auto-apply, auto-send messages or auto-reply on a candidate’s behalf.
  • Auto-update a Job Tracker from email without the candidate’s review and approval.
  • Train external or general models on candidate data without a separate agreement, clear notice and an appropriate legal basis.
“AI can help turn approved information into clearer drafts and preparation ideas. It does not decide suitability for a role, guarantee an outcome or replace your judgement.”
Gael Talent Bridge AI feature principle

Data minimisation and human review

AI features should receive only the minimum relevant text for the requested task. Entries marked private are excluded. Where an AI provider is used, we will assess its role, contract, security, data location and retention controls before launch. We will maintain an activity record covering the feature, input source, time, output state and candidate decision where appropriate.

People remain in control: a candidate can edit, reject or use a manual route instead of an AI-supported feature. Consultants review service material; employers do not receive AI-generated candidate assessments because the service does not create them.

Opting out

AI assistance is optional for candidate-facing drafting and reflection features. You may opt out by choosing the manual route, disabling the relevant feature where available, or contacting info@gaeltalentbridge.ie. Opting out will not prevent us from delivering non-AI parts of a requested service, although a particular optional feature may be unavailable without it.

Before any significant launch

We will complete a Data Protection Impact Assessment before large-scale AI or email processing goes live, and revisit it when the processing changes. The Data Protection Commission’s DPIA guidance describes when and how organisations should assess high-risk processing. We will also test privacy controls, bias risks, security, user comprehension and the approval flow before deployment.

Owner to confirm: the selected AI provider, contractual terms, data-hosting location, retention settings, human-review roles and final opt-out implementation before any AI feature is launched.

CARA voice interview practice

Signed-in candidates granted access by the owner can choose voice practice in Interview Studio without payment or a subscription, with up to three sessions per Dublin calendar day. The browser asks for microphone permission and may process speech through its own speech-recognition service. CARA receives and stores an editable transcript and approximate speaking time, not a raw audio recording. Questions and coaching can be read aloud using browser speech synthesis where supported. CARA analyses interview answers and provides delivery coaching from the transcript and approximate timing; it does not infer an anxiety, stress or medical level. The candidate can remove a voice session’s transcript, report and delivery notes from CARA’s active records. Operational backups may persist under applicable backup-retention controls. Written practice remains available.

Voice practice uses CARA’s configured AI provider for answer analysis and coaching. Its provider terms, data-handling arrangements and any necessary privacy assessment must be reviewed by the owner before enabling voice access for candidates.