Cortex

Cortex is an AI-powered virtual CFO that understands, predicts, and guides SMEs with real-time financial insights—turning data into decisions, not spreadsheets, through voice and intelligence.

Comment

Cortex

The financial intelligence that listens to you.

Inspiration

In Mexico, 75% of SMEs fail before reaching their second year.
Not because they sell too little, but because they don’t understand their own finances.

How many times has a company taken a loan it didn’t need, or invested just when it should have saved? Today, we want to change that.

More than 60% of SMEs lack digital financial management, and 40% rely on personal credit cards to stay afloat. Cortex transforms those scattered numbers into actionable intelligence: it anticipates risks, suggests adjustments, and recommends whether it’s time to save, invest, or seek financing.

Because companies don’t fail from poor sales —
they fail from not understanding themselves.

What it does

Cortex is the first virtual CFO powered by artificial intelligence and voice, designed for Mexican SMEs.

It analyzes a company’s full financial pulse — income, expenses, payroll, suppliers, debts, and cash flow — and in seconds determines its status:

  • Healthy — Can invest and grow.
  • Regular — Needs to optimize resources.
  • Critical — Must adjust expenses or financing.

It also explains why the company is in that state and what actions can improve it. It allows users to simulate financial “What If” scenarios, such as:

[ \text{What happens if I increase prices by } 3\%? \quad \text{What if I reduce expenses by } 5\%? ]

The result is a reasoned, contextual, and conversational response:

“Your projected cash flow remains positive. You can safely invest up to 10%.”

How we built it

We designed an architecture based on microservices, coordinated by an MCP (Master Control Processor) that functions as the central brain of the platform.

The MCP automatically integrates information from Banxico, INEGI, Banorte datasets, and each company’s operational data (income, expenses, payroll, suppliers, debts, and cash flow). It normalizes, cleans, and correlates this data in real time to provide precise answers with complete context.

Model Context Protocol (MCP)

We also integrated a Model Context Protocol (MCP) — a protocol that allows the AI to connect directly with our data sources, tools, and systems in real time. In our case, we use Claude’s MCP (Anthropic), enabling the model not only to generate text but also to dynamically interact with live information, execute actions, and query databases securely and contextually. This transforms AI from something passive into an active agent, capable of making informed decisions and generating personalized responses grounded in real-world data.

Technical Architecture

Module Function Technologies
Data Engine Data extraction, cleaning, and transformation Python, Pandas, Banxico & INEGI APIs
Predictive Core Inflation, exchange rate, and cash flow modeling Prophet (Meta), NumPy
Voice & Reasoning Engine Contextual reasoning and conversational interaction Gemini API
Frontend Gateway Visual interface and conversational mode HTML, CSS, JS, Web Speech API

The data pipelines remove outliers and manage weekly and monthly time series. Predictions are exposed through secure REST endpoints, and each container is isolated to ensure scalability and security.

Conversational Interaction

User interaction is available through both voice and text, allowing a flexible and accessible experience from any device.

The voice mode is activated by a red animated circle that reacts to the user’s rhythm while the system listens, analyzes, and responds.

Interaction Flow

[ \text{Input (voice or text)} \rightarrow \text{Gemini analyzes context} \rightarrow \text{Cortex generates a response} ]

The result is a smooth experience: business owners can speak or type to check their company’s financial status. Cortex responds with strategies, predictions, and explanations based on real data—no need to interpret complex graphs or dashboards.

Challenges we ran into

  • Integrating heterogeneous official sources from Banxico, INEGI, and Banorte datasets (with different structures).
  • Adapting Prophet to non-stationary and volatile financial time series.
  • Synchronizing conversational reasoning and data retrieval in real time.
  • Designing an intuitive interface for users without financial training.

Accomplishments that we're proud of

  • Built the first virtual CFO with conversational AI for Mexican SMEs.
  • Implemented a dockerized, modular, and portable architecture.
  • Integrated financial AI, predictive analytics, and contextual reasoning.
  • Translated complex financial data into clear and actionable insights.

What we learned

We learned to combine three key dimensions of modern knowledge:

[ \text{Conversational Intelligence} + \text{Financial Modeling} + \text{Distributed Architecture} ]

And we realized that clarity and technological empathy are just as valuable as analytical precision. The greatest challenge wasn’t building AI—it was making it understandable and truly useful for those who need it most.

What’s next for Cortex

  • Integration with Banorte’s official SME API for automated diagnostics.
  • New modules for credit risk and real-time cash flow forecasting.
  • A native mobile version with offline voice recognition.
  • Expansion into other sectors: agriculture, retail, and education.

Our vision: for every Mexican SME to have its own virtual CFO —
an ally that listens, analyzes, and guides.

Closing

With Cortex, SMEs can know in seconds whether they have room to invest or need to optimize spending. No spreadsheets. No costly consultants. Just a voice that listens and a mind that guides.

Cortex — the financial intelligence that listens to you.

Built With

Share this project:

Updates