An AI bot that runs the first stakeholder interview on a new casino product idea, so a product manager doesn't have to sit through it — then hands the team a brief worth triaging.
While working on a casino product team, we set up a process for people across the company to bring us their ideas — a dealer's read on the pit, a marketing idea for the lobby, a compliance flag on the loyalty program. The team was never short on ideas.
The bottleneck was time. Every idea needed a product manager to sit down with the person who raised it, research it, and ask the same dozen requirements-gathering questions — before anyone could even decide whether it was worth pursuing. Good ideas sat in a backlog of unanswered questions, not because they weren't worth it, but because nobody had gotten around to asking.
PM-bot conducts that first conversation instead. A stakeholder describes their idea in a couple of sentences, and the bot takes it from there — asking one tailored follow-up question at a time, building on every answer, until it has enough to write up a real summary and business case.
The output is a Markdown file clean enough to become the seed of a spec or a prototype straight away — no separate write-up step for the product team.
The bot also carries stated product-management expertise specific to online casino: player acquisition and retention economics, regulatory and responsible-gambling considerations, game and provider integration, payments and KYC — so its questions probe for what a casino PM actually needs to know, not generic startup boilerplate.
A stakeholder writes a few sentences about what they want to build, fix, or flag.
One question at a time, each one adjusted to what's already been said: audience, problem, business model, risk.
Idea plus business case, written up and ready to read — or download as Markdown.
The more detail shared up front, the faster the idea gets triaged, and the better the team can act on it.
This is the actual PM-bot interface below, running in the same demo/fallback mode the real app uses when no Anthropic API key is configured: a fixed set of five casino-flavored questions and a template-filled summary, so the full flow works with nothing to set up. Drop a real API key into the deployed app and this exact interface switches to genuinely AI-generated, tailored questions and an AI-written business case instead.
Same fallback logic as server/fallbackQuestions.js in the
repo, ported client-side for this page — see
the full source on GitHub
for the AI-generated path and the Express backend behind it.
What you just tried is demo mode exactly as it runs in the real app — the fixed question set and template summary are a deliberate fallback for when no API key is configured, not a separate toy version. The full build is done: connect a real Anthropic API key to the deployed app and the same interface asks genuinely tailored questions and writes an AI-generated business case instead of the fixed one here.
The casino/iGaming persona isn't hardwired in, either — the bot's domain expertise lives entirely in its system prompts, so it can be reconfigured for any industry's product team, not just iGaming.
Next iteration: giving the bot a research skill, so it can pull in initial market and competitor context on its own — instead of relying only on whatever the stakeholder happens to answer.