View all posts in the Guides category
Fifteen customer service role plays for angry customers, de-escalation, billing and refunds, cancellations, tech support, and compliance. Each has an opening line, what good sounds like, and what to score. Five run as free AI practice calls.
Vapi, LiveKit, and Pipecat are infrastructure for building a voice agent. A finished agent already knows how to do a job. Here is how I would decide between them, what building really involves after the demo, and when it makes sense to use both.
Professor Achal Bassamboo needed a final that tested thinking on your feet, not the right answer. Here is how Kellogg ran it at scale: unlabeled situations, a hint market, a poker-face avatar, human grading, and the practices that made an executive MBA cohort willing to take it.
Run system design mock interviews where the AI interviewer scribes your requirements into a shared notepad, watches an architecture board as you build it, probes every box you place, and grades you against the reference design. 7 questions you can start tonight.
Solo casing works up to a point, and that point is the exhibit. A tested ChatGPT and Claude interviewer prompt, the exact places it breaks on a real Kellogg case, how to fix each one, and a five week plan that splits reps between AI and a human partner.
Run full-length AI mock interviews for McKinsey, Bain and BCG where the interviewer shows exhibits on screen, probes your math and pushes back on your recommendation, then grades you against the case's answer key. 29 casebook cases you can start tonight.