Inspiration
From Financial Data to Executive Decisions
Financial statements contain an enormous amount of valuable information, yet transforming that information into actionable business decisions still requires significant manual effort and financial expertise. Small and medium-sized businesses often lack access to experienced financial analysts, making it difficult to identify risks and opportunities hidden within their own data.
Neovantia-CFO was born from a simple question:
Can AI help transform raw financial statements into meaningful executive insights while keeping financial calculations transparent and trustworthy?
Rather than relying exclusively on large language models to interpret spreadsheets, we designed a solution where deterministic financial calculations remain the foundation of every analysis. This approach ensures that key financial indicators are reproducible, auditable, and based on established financial principles, while AI becomes an assistant that explains results rather than inventing them.
AI-Assisted Engineering
Human-led architecture The architecture always remained the authoritative source for implementation decisions.
The software architecture, PRD, technical specification, milestones, and acceptance criteria were designed by the development team.
GPT-5.6
GPT-5.6 served as the engineering advisor. It helped refine the architecture, review technical decisions, improve documentation, and validate the implementation throughout the project.
OpenAI Codex
OpenAI Codex acted as the implementation engineer. Guided by the Architecture Guide, PRD, and Technical Specification, Codex implemented each milestone incrementally. Every milestone was reviewed and validated before moving to the next one.
Why this approach?
Rather than using AI as a simple code generator, we treated it as part of a structured software engineering process. Human developers remained responsible for product vision, architecture, and acceptance testing, while AI accelerated implementation and documentation.
What it does
Neovantia-CFO transforms traditional financial statements into actionable business insights.
The application allows users to:
Upload structured financial statements in Excel format. Validate workbook integrity and required worksheets. Extract financial information automatically. Calculate deterministic financial KPIs. Present the results through an intuitive dashboard. Provide a solid foundation for AI-assisted executive reporting.
Rather than replacing financial analysis with AI, the application combines reliable business rules with a modern architecture designed to support future AI-powered decision assistance.
How we built it
One of the most distinctive aspects of this project is not only the application itself, but the engineering process behind it.
Instead of using AI merely as a coding assistant, we adopted an AI-assisted software engineering workflow.
The development process followed these stages:
- Architecture Design
- Product Requirements Definition (PRD)
- Technical Specification
- Milestone Planning
- AI-Assisted Implementation using OpenAI Codex
- Human Architectural Review
- Acceptance Testing
By separating architecture from implementation, we were able to use Codex as an implementation engineer while retaining full human control over product decisions and software architecture. This approach allowed us to move remarkably fast without sacrificing maintainability or design consistency.
Challenges we ran into
The biggest challenge was time.
The application was conceived, designed, implemented, tested, and prepared for deployment in approximately one working day during the hackathon.
With such limited time, every engineering decision mattered.
Instead of investing hours polishing the user interface, we deliberately focused on building a clean architecture, validating financial workbooks, implementing deterministic KPI calculations, creating a reliable upload pipeline, and establishing a solid foundation for future AI-powered financial analysis.
The result is a functional MVP whose architecture is ready to evolve well beyond the hackathon.
Accomplishments that we're proud of
During the hackathon we are especially proud that we:
Designed a complete software architecture before writing code. Built a functional end-to-end MVP in approximately one day. Used OpenAI Codex as an implementation engineer within a structured development process. Maintained human architectural control throughout the project. Produced a foundation that can evolve into a production-ready AI financial coworker.
What we learned
This project reinforced an important lesson: AI is most powerful when it augments engineering rather than replacing it.
Using Codex as part of a structured engineering workflow dramatically accelerated development while preserving software quality.
We also learned that separating deterministic business logic from AI-generated insights creates systems that are both more reliable and easier to extend. Rather than relying on an LLM to perform financial calculations, the application computes financial metrics using explicit business rules and reserves AI for interpretation, explanation, and executive communication.
What's next for Neovantia-CFO
Neovantia-CFO is only the beginning.
Our long-term vision is an AI Financial Coworker capable of continuously monitoring financial performance, identifying trends, answering executive questions, and helping organizations make better strategic decisions.
The current MVP establishes the architectural foundation for that vision, and the engineering methodology developed during this hackathon will continue to guide future AI-powered products.
Log in or sign up for Devpost to join the conversation.