Navi.

The problem

Mainstream fintech and budgeting apps are designed for users who can read fluent English, navigate forms, and trust a financial app with their data. That excludes a huge population: migrant workers, blue-collar earners, young professionals in early career, and people in financial stress.

For these users, the problems compound: low digital and financial literacy makes text-heavy apps useless; trust gaps (fear of hidden fees and scams) lock people out before they start; fragmented cash-based data means thin or non-existent credit files; regulatory and privacy concerns around SMS and financial data create real adoption barriers; and user fatigue kills any product that demands typing, reading, or dashboard navigation.

The solution

Navi inverts the input model. Instead of forms, users speak ("I spent 12 dirhams on breakfast"), snap a receipt, or let Navi parse SMS transaction alerts. The AI categorizes, logs, and replies conversationally — no dashboards required.

It's positioned explicitly as a budgeting assistant, not a financial advisor. That scope discipline is the product: it keeps Navi safe across jurisdictions, keeps users from over-trusting AI for advice it can't legally give, and keeps the AI's job narrow enough to do well — track expenses, surface patterns, send timely nudges, and answer simple affordability questions.

How it works

Activation: wake word or app open — no typing required. Intake: voice, receipt photo, or SMS read. AI categorization: messy real-world transactions (Walmart purchase split across groceries, household, electronics) are decomposed correctly without manual splitting. Real-time output: a plain-language reply, an optional action, an optional profile update — all served as strict JSON with no markdown or free text outside the contract.

Primary model: GPT-5o-mini, chosen for the cost/latency/safety balance and intentionally constrained by Navi's system prompt to never offer financial advice or assume the user's country. Capabilities live behind a single coordinator: voice logging, receipt OCR, SMS expensing, category prediction, recurring payment detection, daily summaries, weekly insights, spike detection, drift alerts, subscription detection, affordability check, spending simulation, goal impact preview, prioritized bill reminders.

Who it's for

Three end-user personas, all real-world underserved: (1) migrant and low-income workers — managing remittances, irregular income, and limited bank access; (2) young professionals — building first-job financial habits without dashboard fatigue; (3) budget-anxious financial strugglers — drowning in untracked spending and overdue bills.

Business model: B2C-first, with a B2B2C secondary engagement model — banks, remittance partners, and employers offering Navi to their workforce as a financial wellness benefit. Phase 2 expands to deep localization with dialect-tuned voice and culturally-adapted nudges.

Why it matters

Across Navi's three personas the addressable market is expanding rapidly, driven by digital financial tool adoption, embedded finance, and conversational AI. The shape of money advice is shifting from text-heavy dashboards to voice and ambient AI — and the populations that benefit most from that shift are exactly the ones excluded from the previous wave of fintech.

Navi is a thesis bet: inclusive UX (voice-first, low-literacy) plus trust-by-design (bias checks, transparent consent, migrant-safe defaults) plus scope discipline (not financial advice) can reach users that traditional apps can't, and build habits — not just transactions — in the process.

At a glance

Project

Navi

Built by

Aida Michel, Raj, Tatiana, Chiheb & Nidhi

One-liner

Voice-first AI budgeting coach for migrants, blue-collar workers, and the financially anxious — hands-free, low-literacy, trust-by-design.