Parthean

Parthean processes natural language queries about spending patterns and budget status, converting free-form questions into structured financial data queries against connected bank/transaction feeds. The system uses intent recognition to map user questions (e.g., 'how much did I spend on groceries last month?') to transaction category filters and time-range aggregations, returning contextual summaries rather than raw data. This eliminates manual spreadsheet entry by allowing users to ask questions in plain English rather than navigating UI menus or writing formulas.

Parthean analyzes user financial profile (income, spending patterns, debt, goals, risk tolerance) through conversational discovery and generates tailored recommendations for savings, debt payoff, or spending adjustments. The system uses rule-based or LLM-driven reasoning to match recommendations to individual circumstances rather than delivering generic advice, considering factors like income stability, family size, and stated financial goals. Recommendations are delivered conversationally with explanations of the reasoning, making financial guidance accessible to users intimidated by traditional advisor jargon.

Parthean maintains conversation state across multiple user queries, allowing users to ask follow-up questions, refine previous answers, and build on prior context without re-explaining their situation. The system uses session-based memory to track disclosed financial information, stated goals, and previous recommendations, enabling natural dialogue flow. This architectural pattern treats financial planning as an iterative conversation rather than discrete Q&A interactions, reducing cognitive load on users who would otherwise need to repeat information.

Parthean integrates with bank APIs (likely via Plaid, Yodlee, or direct bank connections) to aggregate transaction data from multiple accounts, normalizing merchant names, categorizing transactions, and maintaining a unified view of user financial activity. The system handles OAuth-based authentication to securely access bank data without storing credentials, and periodically syncs new transactions to keep the data current. This aggregation layer abstracts away the complexity of connecting to dozens of different bank APIs, presenting a unified data model to the conversational AI layer.

Parthean automatically categorizes transactions into standard financial categories (groceries, utilities, entertainment, etc.) using merchant name matching, transaction description analysis, and potentially ML-based classification. The system normalizes merchant names across banks (e.g., 'AMZN' and 'Amazon.com' both map to 'Amazon') and applies consistent category rules. Users can refine categories conversationally ('that Amazon purchase was actually a gift, not personal shopping'), and the system learns from corrections to improve future classifications. This eliminates manual categorization friction while maintaining accuracy through user feedback.

Parthean allows users to define financial goals (emergency fund, vacation, down payment) conversationally and tracks progress toward those goals by analyzing spending patterns and savings rate. The system calculates time-to-goal based on current savings velocity and provides conversational updates on progress. Goals are contextualized within the user's overall financial picture, allowing the system to recommend adjustments to spending or savings to accelerate goal achievement. Progress is visualized through conversational summaries rather than charts, making goal tracking accessible without dashboard navigation.

Parthean analyzes user debt (credit cards, loans, student loans) and recommends payoff strategies (avalanche, snowball, or custom) based on interest rates, balances, and user preferences. The system calculates payoff timelines and total interest paid under different strategies, allowing users to compare approaches conversationally. Recommendations account for user circumstances (income stability, other financial goals) and can suggest adjustments to payment amounts or strategy if goals change. The system explains the trade-offs between strategies in plain language, helping users make informed decisions rather than following generic advice.

Parthean analyzes historical spending patterns to identify trends, seasonal variations, and unusual transactions. The system calculates average spending by category, identifies month-to-month variations, and flags transactions that deviate significantly from normal patterns (e.g., unusually large purchase, new merchant category). Anomalies are presented conversationally ('You spent 40% more on dining this month than usual — want to explore why?'), allowing users to understand their spending behavior without manual analysis. This pattern recognition helps users identify budget leaks and understand their financial behavior.

Parthean compares actual spending against user-defined budgets, calculates variances by category, and recommends adjustments when spending deviates significantly from planned amounts. The system identifies categories where users consistently overspend or underspend, and suggests realistic budget adjustments based on historical patterns. Recommendations are delivered conversationally with explanations of the reasoning ('You've budgeted $200 for groceries but spent $280 the last 3 months — should we adjust your budget to $300?'). This approach treats budgets as living documents that adapt to actual behavior rather than fixed constraints.