Building Ball Watching: How We're Using Agentic AI to Automate a Football TV Show

There's a show format that every British football fan knows. Saturday afternoon. A studio full of ex-pros. Goals flashing up on a big screen. Someone shouting "YESSS!" at a scoreline no one expected. It's chaotic, opinionated, and brilliant — and we've spent the last year building an AI version of it.

That product is Ball Watching. And the platform powering it is Aigentique.

Here's how it works, why we built it, and what we're trying to prove.

The Problem We're Solving

Sports media runs on a 24/7 cycle, but production doesn't. Creating live commentary, match analysis, and contextual storytelling at scale requires enormous human resource — studios, analysts, production teams, scheduling. For independent creators and smaller broadcasters, that wall is nearly impossible to climb.

At the same time, the data has never been richer. Live match events, player stats, historical records, form guides, league standings, press conferences, injury news — it's all out there, structured and available, just waiting to be turned into something a fan actually wants to watch.

The gap between that data and a finished piece of content is exactly what Aigentique is built to close.

What Is Aigentique?

Aigentique is our agentic AI platform — a modular intelligence and orchestration system that ingests live and historical data and turns it into automated, narrative-driven content.

It's not a single tool or a chatbot. It's a full production pipeline with seven stages:

Data → Signals → Intent → Segments → Script → Rendering → Delivery

At the foundation is a medallion data lake on AWS (S3 Bronze, Silver, Gold layers processed by 14 Glue PySpark ETL jobs), combined with DynamoDB for live match state — polling every minute, detecting changes via MD5 hashes, and routing events like goals, red cards, and pressure shifts through Kafka.

Above that sits two core engines:

• The Orchestration Layer — which controls when things happen and where they go. It detects signals, manages show scheduling across 9 dynamic states (from PRE_MATCH through to OffAir), and delivers content via WebSocket in real time.
• The Intelligence Layer — which controls what to say and how to say it. LangGraph agents running on ECS Fargate generate personality-driven commentary via Amazon Bedrock (Claude Sonnet 4.5). A mode-aware showrunner selects the right prompt template based on match context. Script memory prevents repetition across invocations.

The two layers meet at Intent — the handoff point where "something happened" becomes "here's what we should talk about and why."

The entire thing runs unattended. No human has to press play.

What Is Ball Watching?

Ball Watching is our flagship product built on Aigentique. Think Sky Soccer Saturday or BBC Final Score — but automated, live, and AI-led.

It's a live-first broadcast covering 8 English football leagues simultaneously: Premier League, Championship, League One, League Two, National League, the League Cup, EFL Trophy, and FA Cup.

The show runs across 9 channel modes that change automatically based on the football calendar — pre-match build-up, live coverage, post-match analysis, matchday round-ups, news mode, podcast format, and off-air. Pre-match and post-match blocks are 30 minutes each, auto-generated and scheduled without manual intervention.

The frontend (built in Next.js 16 / React 19) delivers a live ticker, vidiprinter, pressure index charts, and contextual advertising — all updated via WebSocket in a single composite payload per fixture that includes state, scores, pressure delta, elapsed time with stoppage time formatting, and events.

The name itself tells you the tone. "Ball watching" is a football sin — the cardinal error of the defender who's admiring play instead of doing their job. We've leaned into that: a show with big opinions, suspect analysis, and the kind of banter that happens when you care too much about results that don't actually affect your life.

Why We Built It This Way

We made a deliberate choice early on: build a platform, not a show.

Ball Watching is an expression of Aigentique, not special-cased logic buried inside it. Our second product, Full Kit Shankers — an entertainment-first, comedic football show — runs on the same underlying infrastructure. Same pipeline, same agents, same data. Different prompt templates, different presenter tone, different editorial guardrails.

That matters because it means the system scales horizontally. Adding a new show, league, or sport is a configuration change, not a rewrite.

We also made a deliberate choice about how AI fits into this. The agents in Aigentique are not free-running generators — they're invoked deliberately by the Orchestration Layer with scoped context, iteration caps, prompt injection protection, and clear data contracts. This is what separates a reliable automated broadcast from a hallucinating chatbot. The data architecture enforces it: live state and historical context are separated by design, so the AI always knows what's real and current versus what's background knowledge.

Where We Are Now

We are still in active development and our Phase 2 completed in April 2026. The platform now supports pre-match and post-match automation, mode-aware prompt selection, a fully upgraded WebSocket delivery system, and three new frontend layouts. There are hundreds of unit tests covering infrastructure, agents, ETL, and shared modules.

The tech stack is AWS-native end-to-end: CDK (Python 3.12), DynamoDB, S3, Athena, Bedrock, Kafka via Confluent Cloud, ECS Fargate, EventBridge, CloudFront — with a Next.js / React / TypeScript frontend.

Phase 3 is already scoped: digital human presenters via WebGL avatars and ultimately Unreal Engine MetaHuman with Pixel Streaming. The goal is full vertical integration — AI that doesn't just write the words, but delivers them on screen through a rendered presenter.

The Bigger Picture

What we're building with Aigentique is a proof of concept for something much larger: a fully autonomous sports broadcasting network, where AI journalists, commentators, and storytellers cover football — and eventually all sports — in real time, at scale, across every digital platform.

Ball Watching is the first show. It's the thing we point to and say: this is what the platform can do.

There's a lot more to come.

Triple P Digital is a digital product studio building AI-powered products and platforms. Aigentique Sports and Ball Watching are in active development.