Skip to main content
Ask

Your findings,
in plain English.

Ask isn't a chatbot. It's a semantic layer over every insight and recommendation that discovery has produced — type a question and surface the findings that actually answer it, with sources.

app.decisionbox.io/ask
Ask your insights — question and grounded answerAsk your insights — continued answer with cited sources

A real Ask session — the question, the grounded answer, and the cited source insights. Nothing invented, everything traceable.

Not a chatbot

Retrieval, not invention

A chatbot answer
"Based on your data, cancellations appear to be elevated. You may want to look into what's causing guests to cancel. Would you like me to generate a report?"
An Ask answer
3 matching findings, all from a discovery run you can trace.
Source
Run #142
Kind
Insights
Latency
0.4s
Every result is a validated insight — not a paraphrase, not a guess.
Sources on every answer
What makes it different

Built to access, not invent

Grounded in real findings

Ask only returns insights and recommendations that discovery has actually produced. Nothing invented, nothing pulled from context.

Semantic, not keyword

Ask for "retention risks" and it finds findings about churn, cancellation, and repeat-booking rate — even if none of those words match literally.

Sources on every result

Every answer links back to the discovery run, the severity, the affected count, and the original validated evidence.

Insights and recommendations

One search surface for both. Ask "what should we prioritize?" and it blends high-impact findings with their matching actions.

Ask lives on top of discovery

Every question Ask answers is backed by an insight or recommendation that a discovery run already produced and verified. There's no separate model, no training your team has to do — as soon as discovery finishes, the findings are searchable in plain English.

Try it in two minutes

Clone the repo, run docker compose up, and point it at your warehouse.