The challenge
Mocki had to do something scripted practice tools cannot: hold an interview that responds to what the candidate actually says. A fixed question list teaches candidates to recite answers; a real interview follows up, probes and changes direction.
It also had to work in a browser with no install, because the moment of need is the night before an interview, and it had to feel human enough that practising in it is not embarrassing.
An adaptive interview loop
We built the interview as an LLM-driven conversation rather than a questionnaire. Each response feeds the next question, so the session follows the candidate the way an interviewer would.
The workflow logic controls pacing and structure around that loop, so an interview stays coherent from opening question to close instead of drifting.
Built browser-first on Bubble.io
The whole application is a Bubble.io build — accounts, sessions, conversation state and interface — delivered in the browser with nothing to download.
Choosing Bubble.io meant the conversational experience could be tuned iteratively, which matters for a product whose quality is defined by how the interaction feels rather than by a feature list.
Coaching, not just questioning
The experience is framed around building confidence and sharpening communication, so the product is positioned as coaching a candidate through practice rather than testing them.
That framing shaped the interface: calm, focused, and free of anything that turns a practice session into an exam.
The outcome
Mocki is live as an interactive AI interview coach that runs adaptive, human-like practice interviews in the browser.
It is one of three AI and LLM products in our portfolio, alongside Tylo AI and InkGenX, and the clearest example of our conversational AI work.