FLOYD – Voice-Based AI Assistant for Football Decisions.

FLOYD was built as a focused mobile application that allows users to interact with football rules through voice. Instead of searching through rulebooks or relying on manual judgment, users can simply speak and receive instant guidance.

The idea was to create a tool that feels natural to use during discussions, practice sessions, or casual matches where quick decisions are needed.

Where the Idea Came From

The client’s goal was to build a simple but highly functional product that solves a very specific problem.

  • They wanted:
  • A voice-based assistant that understands football scenarios.
  • Accurate responses based on official rules
  • A fast interaction flow without complex navigation
  • A subscription-based model for monetization

The focus was not on adding too many features, but on making one core feature work reliably.

The Problem That Made This Project Hard

Building a voice AI is one challenge. Building one that is actually right about football rules is a different problem entirely.

Football rules can be detailed and sometimes open to interpretation. Translating these into a system that provides consistent answers required careful planning.

  • Key difficulties included:
  • Handling different variations of the same scenario
  • Accurate responses based on official rules
  • Ensuring responses remain clear and are not confusing
  • Avoiding incorrect or misleading decisions

How Our Team Approached the Problem

Our AI engineers, designers, and developers worked closely from the start rather than treating each phase in isolation. The process involved continuous refinement to ensure the assistant responded reliably and predictably.

Technologies That Power Floyd

React Native

Node.JS

OpenAI API

Firebase

Docker

What the Client Got at the End of This Project

Floyd is live on both the App Store and Google Play. It does exactly what the client set out to build, and it does it reliably. Users activate it with a word, describe what happened on the pitch, and get a ruling they can actually trust.

The subscription model is running. The AI handles the scenarios it was trained for with consistent accuracy. The codebase is clean, documented, and ready to scale as the user base grows.

Got an AI concept that needs the right team to build it? Tell us what you have in mind.