🤖 AI

FinAgent: An Agentic AI Framework Integrating Personal Finance and Nutrition Planning

This paper presents FinAgent, a modular multi-agent AI framework that dynamically integrates personal finance, health data, and real-time food prices to generate affordable, nutritionally adequate meal plans, demonstrating a 12–18% cost reduction and over 95% nutrient adequacy in a Saudi household case study.

Original authors: Toqeer Ali Syed, Abdulaziz Alshahrani, Ali Ullah, Ali Akarma, Sohail Khan, Muhammad Nauman, Salman Jan

Published 2026-03-19

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Toqeer Ali Syed, Abdulaziz Alshahrani, Ali Ullah, Ali Akarma, Sohail Khan, Muhammad Nauman, Salman Jan

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are the captain of a small ship (your household). You have a limited amount of fuel (your budget), a crew with specific dietary needs (your family's health), and the ocean is constantly changing with unpredictable storms and price spikes (the fluctuating cost of food).

Most current tools are like a static map. They tell you, "Here is a route to take," but if a storm hits or a bridge closes, the map doesn't update. You're left stuck or forced to pay extra to find a new way.

FinAgent is different. It's not just a map; it's an autonomous, super-smart co-pilot that never sleeps.

Here is how this "co-pilot" works, broken down into simple concepts:

1. The Team of Digital Specialists (The Agents)

Instead of one giant brain trying to do everything, FinAgent uses a team of specialized digital assistants (agents) who hold hands and share notes. Think of them as a dream team in a kitchen:

The Accountant: Keeps a close eye on your wallet. It knows exactly how much money you have left after paying rent and bills.
The Nutritionist: Makes sure the crew gets enough vitamins, protein, and iron to stay healthy.
The Price Watcher: This is the radar. It constantly scans the "market ocean" for price changes. If the price of chicken suddenly jumps 20%, this agent spots it immediately.
The Cultural Chef: Ensures the food fits your family's rules (like Halal requirements) and tastes (no one likes eating what they hate).
The Substitution Expert: This is the magic trickster. If the Accountant says "Chicken is too expensive" and the Nutritionist says "We need protein," this expert instantly suggests, "Let's swap chicken for lentils or sardines!" It finds the perfect replacement that keeps the meal healthy but cheaper.

2. How It Solves the Puzzle

The system treats your weekly meal plan like a giant, complex Lego puzzle.

The Goal: Build a tower that is tall enough (nutritious) but doesn't use too many blocks (money).
The Problem: Sometimes, the price of a specific Lego brick (food item) changes overnight.
The Solution: The moment the Price Watcher sees a change, the whole team huddles up. They don't panic; they just rearrange the Lego bricks. They swap an expensive brick for a cheaper one that fits the same spot, ensuring the tower stays strong and the budget stays safe.

3. Real-World Results (The Test Drive)

The researchers tested this system on a typical Saudi family of four. Here is what happened compared to using a standard, static meal plan:

Saving Money: The system saved the family 12% to 18% on their grocery bill. That's like getting a free week of groceries every month just by being smarter about what you buy.
Staying Healthy: Even with the cheaper food, the family still got 95% or more of the nutrients they needed. They didn't have to choose between eating well and saving money.
Weathering Storms: When prices for things like chicken or rice jumped by 20% or 30%, the system automatically adjusted the menu. A static plan would have broken the budget, but FinAgent just swapped ingredients and kept going.

4. Why This Matters

For a long time, people have had to choose: "Do I eat healthy, or do I save money?" Usually, low-income families end up eating cheap, unhealthy food, which leads to health problems later.

FinAgent proves you don't have to choose. By using AI to act as a personal financial and health guardian, it makes it possible to eat a high-quality, culturally appropriate diet without breaking the bank. It turns the stressful task of "What's for dinner?" into a solved equation, giving families more peace of mind and more money in their pockets.

In short: FinAgent is like having a personal chef, a financial advisor, and a grocery shopper rolled into one, working together 24/7 to make sure your family eats well, stays healthy, and saves money, no matter how crazy the market gets.

1. Problem Statement

The paper addresses the critical challenge faced by middle-income households: balancing limited financial budgets with complex nutritional demands in a volatile market environment.

The Gap: Existing tools are siloed. Financial apps (e.g., Mint, YNAB) track expenses but ignore nutrition, while diet planners suggest meals without considering real-time price fluctuations or household budget constraints.
The Consequence: This disconnect leads to suboptimal dietary choices, increased long-term health risks (especially for low-to-middle-income families), and failure to meet Sustainable Development Goals (SDG-2: Zero Hunger, SDG-3: Good Health).
Specific Context: The study focuses on the Saudi Arabian market, where food prices fluctuate, and cultural/religious constraints (e.g., Halal, Ramadan) must be strictly adhered to.

2. Methodology: The FinAgent Framework

The authors propose FinAgent, a modular, multi-agent AI system designed to jointly optimize food budgeting and nutritional adequacy. Unlike static planners, FinAgent is "agentic," meaning it possesses autonomy, planning capabilities, and the ability to react to environmental changes.

A. System Architecture

The framework utilizes a Multi-Agent System (MAS) where specialized agents coordinate via a shared knowledge base:

Budget Agent: Calculates the weekly disposable food budget ( $B_w$ ) based on income and fixed expenditures.
Price Monitor: Tracks real-time supermarket prices (via APIs and scrapers) and detects "price shocks" (defined as deviations $>10\%$ ).
Nutrition Agent: Ensures macro- and micronutrient adequacy based on WHO/FAO/USDA guidelines.
Health Personalizer: Adjusts nutrient targets dynamically based on medical conditions (e.g., diabetes, allergies, deficiencies).
Cultural & Preference Agent: Enforces dietary rules (Halal, vegetarian, local cuisine preferences).
Substitution Agent: Uses a substitution graph to identify alternative food items that maintain nutritional quality when prices spike.
Procurement Agent: Converts optimized plans into actionable shopping lists.
Explainer Agent: Provides transparent reasoning for decisions to build user trust.

B. Mathematical Optimization Model

The core planning engine formulates meal planning as a Linear Programming (LP) problem:

Objective: Minimize total cost ( $Z = \sum p_j x_j$ ).
Constraints:
- Nutritional: $\sum c_{j,n} x_j \geq R_n$ (Total nutrient intake must meet household requirements).
- Budgetary: $\sum p_j x_j \leq B_w$ (Total cost must not exceed the weekly budget).
- Non-negativity: $x_j \geq 0$ .
Adaptivity: When the Price Monitor detects a shock ( $|p_{j,t} - p_{j,t-1}| / p_{j,t-1} > \tau$ ), the system triggers an automatic re-optimization, utilizing the Substitution Agent to swap items while solving the LP again.

C. Implementation Details

Tech Stack: Python (PuLP, OR-Tools for optimization), Multi-agent LLM coordination, Relational databases for structured data, and a lightweight web UI.
Data Sources: Real-time price feeds from major Saudi retailers (Panda, Carrefour, Lulu), authoritative nutritional databases (USDA, FAO), and synthetic household profiles based on World Bank/CEIC distributions.

3. Key Contributions

Integration of Finance and Nutrition: First framework to dynamically couple household budgeting with nutritional optimization in real-time, addressing the "affordability vs. health" trade-off.
Agentic Adaptivity: Moves beyond static meal planning by introducing autonomous agents that detect market volatility and re-plan meals instantly without user intervention.
Cultural & Health Personalization: Explicitly models cultural constraints (Halal, Ramadan) and medical needs (diabetes, allergies) within the optimization loop, ensuring plans are not just cheap and healthy, but also feasible and safe.
Substitution Graph Logic: Introduces a mechanism to maintain nutritional integrity even when specific ingredients become prohibitively expensive, by mathematically mapping nutritional equivalents.

4. Results and Performance

The system was evaluated through synthetic simulations (200 households, 30 runs) and a real-world 4-week case study with a Saudi household.

Cost Efficiency:
- Achieved a 12–18% reduction in weekly grocery costs compared to fixed/static menus.
- In the real-world case study, the system reduced costs by 17% (from ~2,000 SAR to ~1,660 SAR/month for a 4-person family).
Nutritional Adequacy:
- Maintained >95% nutrient adequacy (including critical nutrients like Iron, Vitamin D, and Calcium) across all scenarios.
- Outperformed manual planning (88% adequacy) and static optimization (92% adequacy).
Resilience to Price Shocks:
- Successfully handled price fluctuations of ±20% to ±30% (e.g., a 20% increase in chicken prices) by automatically substituting with lentils or sardines, keeping the plan within budget.
- Static plans failed to stay within budget under similar shocks.
Ablation Studies:
- Removing the Price Monitor increased costs by 9%.
- Disabling the Health Personalizer dropped Vitamin D adequacy to 70%.
- Removing the Preference Agent reduced user satisfaction scores significantly (3.2 vs. 4.4).

5. Significance and Future Work

SDG Alignment: The framework directly supports SDG-2 (Zero Hunger) and SDG-3 (Good Health and Well-being) by making nutritious diets accessible to middle-income families despite inflation.
Scalability: The modular, API-driven architecture allows for easy integration with loyalty programs, government subsidies, and school meal programs.
Ethical Considerations: The authors highlight the need for privacy-preserving techniques (on-device calculation, differential privacy) and bias auditing in food ontologies to ensure fair recommendations.
Future Directions: Large-scale deployment, long-term compliance studies, and integration with broader economic ecosystems (subsidies, supply chain data).

Conclusion: FinAgent demonstrates that Agentic AI can effectively solve complex, multi-constraint household problems. By automating the trade-off between cost, nutrition, and cultural preferences, it offers a viable, scalable solution for sustainable diet planning in volatile economic environments.