How we use AI — and how we don't
Last updated 18 July 2026. Honesty about AI is a product feature here, not a legal chore.
What uses AI today
- The tuition coach helps you shape your own experiences into clear STAR stories. It uses Anthropic's Claude, called only from our server for your own signed-in session. It asks probing questions and structures your words — it is barred from inventing experience you don't have, and it flags any claim your answers don't support.
- Scored mock interviews generate questions for your role and score each answer against a fixed behavioural rubric (structure, specificity, relevance, evidence, delivery). Also Anthropic's Claude, server-side only. Feedback is restricted to that rubric — it never comments on your emotions, confidence or personality.
- Speech-to-text in spoken practice uses your browser's built-in speech recognition. Audio is processed by your browser vendor, not by us; transcripts stay on your device.
What doesn't use AI
- Your programme is built by a deterministic engine — same inputs, same plan, every time.
- The questions to expect are selected by transparent rules from researched question banks, filtered by your role, stages and region.
- Delivery feedback (words per minute, filler count, length) is simple counting, on your device.
How your material is handled
When you use the coach or a scored mock, the relevant context — your role and company, the stories you're building, and your CV if you've added one — is sent to Anthropic only for that request, only for your own session, and only when you're signed in. The key that authorises those calls never touches your browser; it lives as a server-side secret. Calls are identity-minimised (we don't send your name or contact details), and requests are capped per day to keep the coach a practice aid, not a crutch.
What we refuse permanently
- No live-interview assistance. Intervooh will never feed you answers during a real interview.
- No fabrication. It will not invent experience, projects or qualifications for you.
- No emotion or personality inference. We do not analyse your face, read your "confidence", or score your personality — and EU law now prohibits emotion-recognition AI in recruitment contexts. We measure only what you can control: structure, specificity, relevance, evidence and delivery mechanics.
Limitations, stated plainly
- Question banks are researched and representative, but no tool can know exactly what your interviewer will ask.
- Speech recognition mishears sometimes, especially with background noise — the transcript is a practice aid, not a judgement.
- Preparation improves your odds; nothing guarantees an offer.