An explainable Vedic astrology product for web and mobile — every reading links back to the calculation, the classical rule and the reasoning behind it.
Astrology products hand down conclusions. AstroAI separates what is calculated from what is interpreted.
People are told what their chart means without being shown why. Settings are hidden, apps contradict each other, and AI astrologers answer fluently without grounding.
Treat explainability as the product: a reasoning chain, provenance labels on every statement, and contradictions shown side by side.
A responsive website and a mobile app on one design system, with a “Why?” on every reading and a North Indian chart at the centre.
How might we help people understand what their birth chart says — and why — without asking them to believe it?
From the product definition. Each is a hypothesis for the research plan, not a finding.
A “Why?” on every interpretation.
Method and settings shown up front.
Calm language; no remedies commerce.
Supports and challenges side by side.
Progressive disclosure, from simple to technical.
A grounded Ask that lists every factor it used.
A living dasha timeline.
Confidence and sensitivity badges.
Proto-personas that set the depth system — Simple, Detailed and Technical. To be validated in interviews.
Help me understand what my chart actually says, and on what basis — without asking me to believe anything.
Show me why two apps disagree, and what my current dasha actually activates.
Give me an engine I can trust and a rule trace I can check against the texts.
Explain my sister’s chart to me honestly.
Apps state outcomes; nobody can evaluate or disagree with them.
Every reading opens a short sheet: where it is, the rule, what it means for you, and one caution — closing it returns you to the same place.
The rule behind any reading is three taps away, without losing context.
Facts, tradition and opinion read the same, so everything sounds equally certain.
Four voices, four colours: calculated, classical rule, interpretation and AI summary.
People can tell a computed fact from an interpretation at a glance.
Products show only the flattering side of mixed indicators.
“What supports it, and what doesn’t” lists challenging factors next to supporting ones.
Honesty about disagreement should build trust — a hypothesis to test.
AI astrologers produce fluent answers with no link to the chart.
Ask retrieves chart factors first, removes unsupported claims and shows a “grounded in” list.
Every claim in an answer traces back to a factor the user can inspect.
Final designs — Saffron Sunrise · Free Kundli and Dasha on the web; Today, Kundli, Why?, Dasha, Ask and Life timeline in the app
This looks like a SaaS dashboard.
Neutral surfaces, serif interpretations, mono data. Strong on structure — but it read as enterprise software.
A familiar kundli-page layout for web and an Airbnb-style app, then five full directions with identical content to compare.
Warm light surfaces, one saffron accent and rounded tinted tiles — one token-based system for web and app.
Outcomes stay empty until they are measured.
The first round had the right reasoning model, but looking like enterprise software undermined trust.
One idea per screen made the “why” easier to find.
Five directions with identical content made the choice quick and concrete.
A trusted local marketplace for construction and home services.