A decision-making AI that weighs your factors, not just yours.
Ryddel turns hard decisions into structured analysis. Users define the factors that matter to them, weight them by importance, and let the AI surface the option that fits their actual values — not a generic web result. We built the reasoning engine, the mobile app, and the web product.
Decision fatigue is real — and most decision-making tools either oversimplify (pros/cons lists) or overcomplicate (spreadsheet scoring models nobody maintains).
Decision fatigue is real — and most decision-making tools either oversimplify (pros/cons lists) or overcomplicate (spreadsheet scoring models nobody maintains). Ryddel's founders believed there was a gap between 'flip a coin' and 'hire a consultant.'
The core challenge: the AI had to feel like it understood context, not just criteria. 'Should I take this job?' requires weighting salary, growth, culture, commute, and personal values differently for every user. A generic scoring model fails immediately.
We needed to ship a web app, an iOS app, and an Android app in 12 weeks — and the AI reasoning layer had to be robust enough for high-stakes decisions.
Structure for the unstructured.
We ran decision-mapping workshops with 28 people facing real decisions (job changes, moves, major purchases). The pattern: people know their factors but not their weights. The UX had to make weight-setting feel natural — sliders and emoji-backed importance scales outperformed numeric inputs by 3x.
Reasoning at the speed of thought.
The web app is Next.js with a real-time reasoning UI that streams analysis as GPT-4 processes each option. Mobile is React Native. The backend persists decision trees, factor weights, and historical choices — so the AI learns your preferences over time.
What ships in the product.
Factor-weighted analysis
Define your criteria, set importance weights, and the AI scores every option against your specific values — not generic rankings.
Streaming AI reasoning
Watch the AI think in real time. Each step of the analysis appears as a human-readable rationale, not a black-box score.
Decision history
Every decision is saved. Revisit past choices, see how the AI reasoned, and compare outcomes to predictions.
Collaborative decisions
Share a decision with a partner or team. Each person sets their own weights — the app shows where you agree and disagree.
Templates library
Pre-built decision templates for job changes, city moves, vendor selection, and 40+ other common decision types.
Outcome tracking
Mark decisions as good/bad 30, 90, and 180 days later. The AI learns what factors you systematically over- or under-weight.
Devil's advocate mode
After reaching a recommendation, the AI argues the opposite case — stress-testing your reasoning before you commit.
Export to PDF
Full decision report with factor weights, option scores, reasoning summary, and recommendation — shareable with stakeholders.
Privacy-first storage
Decisions are E2E encrypted. Ryddel's team cannot read your decision content — only anonymised aggregates power future model training.
Production-grade from week one.
We chose the boring, battle-tested options where it mattered and innovated where it gave the product a real edge.
I used Ryddel to decide whether to leave my job of 8 years. The AI surfaced a factor I hadn't consciously weighted — and I made a decision I've never regretted.
What it took.
- Streaming AI that doesn't feel slow. GPT-4 streaming tokens arrive at variable speeds. We built a typewriter renderer that maintains readable pacing — buffering fast bursts and padding slow ones — so it always feels like thinking, not lag.
- Factor weights that feel natural. Numeric sliders felt like homework. We A/B tested emoji-anchored sliders, plain sliders, and drag-ranking. Emoji sliders won on both completion rate and user confidence.
- Collaborative decisions with disagreement. When two people weight factors differently, showing a single score is misleading. We built a disagreement heatmap that surfaces specifically where values diverge — a feature users said was more valuable than the recommendation itself.
- Decision quality feedback loop. Users need to see whether past AI recommendations were good. We built a 30/90/180-day check-in loop with streak incentives — 60% complete at least one follow-up, generating the outcome data that trains the model.
What changed.
A walk through the product.
Key screens, detailed.
A closer look at the core user flows built for this product.
Building something like Ryddel?
Ryddel turns hard decisions into structured analysis. Send us a brief and we'll scope it within 24 hours.