Navorina

From an assistant to a financial operating system

Navorina is an AI-native workspace for people whose money lives across banks, brokers, crypto services and business tools. It combines financial intelligence with the operational work of getting paid, planning, moving money and closing the month.

The project began in November 2025 as an offline-first voice assistant. Financial questions, calculators and portfolio tools gradually exposed a larger problem: an assistant could explain a concept, but it could not give a trustworthy answer without a durable picture of the person’s actual finances.

I expanded that experiment into a working product with a Financial Mirror, accounts, budgets, transactions, invoicing, payments, portfolio context, briefs and a contextual assistant. I owned the product direction, interaction model, visual system and implementation.

Where the project started

The first product was a conversational interface with local speech recognition, local language models and graceful browser fallbacks. It proved that a private assistant could remain useful without depending on one cloud provider.

The next iterations added financial calculators, portfolio tracking, alerts and reports. Individually, those features worked. Together, they still behaved like a toolbox. Each answer began with new inputs, and no screen owned the financial truth over time.

Turning point

The missing feature was not another calculator. It was a financial state the product could remember, explain and safely act on.

The balance was not the answer

Financially active people rarely keep money in one place. Income may arrive through a business account, savings sit in another currency, investments live with a broker, crypto sits on an exchange and taxes remain an obligation outside every visible balance.

This fragmentation turns basic questions into reconciliation work. What changed this week? How much is actually available after upcoming payments and tax? Where is the risk? A dashboard can show totals and still leave those questions unanswered.

AI creates another risk. Without a governed source of truth, a fluent answer can be more dangerous than no answer. Navorina therefore had to separate calculation from explanation and advice from execution.

Research through product evolution

Navorina did not begin with a formal research programme. The early evidence came from building and using successive versions, documenting failure modes and comparing what each new feature could answer against what remained unknown.

The repository history shows the shift clearly: a general assistant became financial tools, then a persistent Financial Mirror, then invoicing and month-close workflows, and finally provider-neutral payment and AI contracts. Each stage reduced one kind of ambiguity and exposed the next.

1. 01
   Conversation

   A private assistant made financial questions easier to ask, but did not own reliable context.

2. 02
   Tools

   Calculators and portfolio views answered isolated questions without preserving continuity.

3. 03
   Financial Mirror

   Accounts, transactions, budgets and snapshots created one rebuildable financial truth.

4. 04
   Operating system

   Invoices, payments, reconciliation and month close connected understanding to action.

Choosing a narrower market entry

The product can serve personal finance, investing and business operations, but presenting all of that at once makes the value difficult to understand. The current beachhead is financially active adults aged 25–45 with income across several services who say, “I don’t understand what is happening with my money.”

This is a market hypothesis, not a reported result. The validation plan starts with people who use at least three financial sources and tests whether a truthful weekly brief becomes a four-week habit.

Research before scale

The research plan is designed to replace broad enthusiasm with behavioural evidence. Recruiting starts with people who keep money in at least three places. The interview begins with the last real financial decision, not a feature list: which services they opened, what they calculated manually, what remained uncertain and what they did next.

A prototype test then asks each participant to build a first financial state and interpret one weekly verdict. A four-week diary study measures whether the brief is opened, understood and followed by an action. Upgrade and cancellation interviews test which clarity moment is valuable enough to pay for and what remained missing.

1. 01
   10 problem interviews

   Reconstruct the last reconciliation task and quantify tools, time, uncertainty and consequence.

2. 02
   First-state usability test

   Observe whether a person can add sources, understand the result and identify what is still missing.

3. 03
   Four-week cohort

   Track brief opens, verdicts consumed, actions taken and returns without reminders.

4. 04
   Price and cancellation evidence

   Test Pro at $29 with a real checkout, then ask what clarity was worth paying for or absent.

Two product faces, one source of truth

I separated the product into Navorina Intelligence and Navorina OS. Intelligence is the market-facing layer: net-worth direction, weekly briefs, risk, portfolio context and the assistant. The OS contains the deeper operating workflows: accounts, transactions, invoicing, payments, accounting and month close.

Both surfaces use the same Financial Mirror. Income, expenses, categories, balances, snapshots and multi-currency rules belong to one workspace rather than being recalculated independently by every feature.

This structure lets the product start with an understandable promise while retaining the operational depth needed for long-term value. The assistant is an interface to the system; it is not a separate product with its own version of the truth.

Rules calculate. AI explains.

Deterministic rules produce balances, forecasts, tax reserves, health states and reconciliation facts. AI receives bounded context and turns those facts into an explanation: what changed, why it matters and which workflow can resolve it.

The boundary is explicit. Generated narrative cannot mutate a ledger, submit a payout or confirm an invoice. A money action moves through capability check, input, review, submit, durable status and reconciliation. If a live rail is unavailable, the interface fails closed instead of showing a synthetic success.

Trust principle

A confident sentence is never allowed to become a financial fact.

Designing complete financial journeys

The critical journey begins with sign-up and a first financial state, then continues through a real task: create a budget or invoice, save it, return after reload, receive or send when a rail is available and reconcile the outcome.

This changed how I evaluated screens. A successful form was not enough. The created entity had to survive a new session, appear in the right financial period and explain its effect on cash flow. Payment states had to remain legible through pending, completed, rejected, returned and interrupted cases.

The same orchestration contract now supports dashboard actions and deeper Nucleus workflows. Providers remain replaceable; the interface depends on capabilities and durable states rather than a provider brand.

Turning clarity into a business

The commercial model follows the product architecture. A free tier proves the first financial picture. Pro sells recurring intelligence. Plus sells operational depth and automation. Team extends the same workspace to shared responsibility.

The prices below are planning hypotheses, not current reported revenue. They are designed to test whether clarity can support a software margin before Navorina takes on the cost and regulation of deeper financial rails.

Revenue expands only when the product earns it

SaaS is the first revenue layer. Payment fees, embedded APIs and partner revenue from FX or interchange come later, after retention and regulated execution are proven. The product can therefore validate demand without holding customer funds or pretending that partner infrastructure already exists.

Unit economics guardrails

Pro is viable only if AI remains a controlled software cost rather than an open-ended service bill. The model targets gross margin above 80%, AI cost below 10–20% of Pro revenue, payback below three months and LTV:CAC above 3:1 before any paid acquisition begins. Founder-led discovery keeps early CAC close to zero while the retention loop is still being proven.

A growth scenario, not a promise

The first 100 users are a high-touch validation cohort. The internal goal of 50 paying Pro customers is intentionally a stretch test and should not be treated as a normal 50% conversion forecast. A broader model uses the planned 3–5% Free-to-Pro range once acquisition expands beyond the founder’s network.

At 10,000 active free users, Pro alone would produce the scenarios below. They exclude Plus and Team upgrades, payment fees, churn, tax and partner revenue, so the model exposes the core subscription engine without stacking unproven assumptions.

The model becomes more attractive only if Plus converts users who already rely on invoicing, month close and exports. That expansion should be measured as retained operational value, not assumed in the forecast.

A working system with explicit gaps

The product already includes the Financial Mirror, invoicing, transaction and account workflows, month close, briefs, alerts, contextual AI, payment orchestration, non-custodial crypto flows and a paywall framework. Core journeys are protected by automated contracts and browser tests.

Implemented does not mean live. Production Stripe cutover, selected fiat providers, controlled crypto pilots and founder-cohort retention remain open gates. The case keeps those boundaries visible because product credibility depends on separating a working workflow from a regulated service that is ready for real money.

Where the product should go next

The next milestone is not another module. It is evidence that the narrow promise works. The roadmap stays vertical: prove clarity, charge for the repeated value, then expand into operational depth and regulated execution.

1. 0–3
   Prove the habit

   Recruit 10 qualified users, run interviews and first-state tests, and deliver four consecutive weekly briefs.

2. 3–6
   Prove revenue

   Cut Stripe over to production, sell Pro at $29 and measure activation, retention, conversion, AI cost and cancellation reasons.

3. 6–18
   Grow the clarity wedge

   Reach the first 1,000 users through useful financial tools, founder content, referrals and shareable briefs—without paid ads before retention.

4. 18–36
   Expand into operations

   Move retained users into Plus through invoicing, close calendar, expected-versus-received cash flow and accountant exports.

5. 36+
   Open the infrastructure layer

   Add partner-mediated payment rails, embedded APIs and team controls only after compliance, reconciliation and support are repeatable.

The strategic priority remains Navorina Intelligence. Bank aggregation, exchange connections and payment infrastructure can deepen the product, but none should distract from the first proof: people return because Navorina makes their financial state easier to understand and act on.

What I learned

AI becomes more useful when its authority becomes narrower. Separating deterministic financial truth from generated explanation produced a system that is easier to trust and test.

A complete journey creates more product value than a large feature inventory. Save, restore, review and reconcile states shaped Navorina more than the ideal dashboard did.

The most durable product advantage is not a chatbot or a bank account alone. It is the combination of workspace truth, decision history, recurring financial rituals and the infrastructure that can carry an approved action safely to completion.