AI at ScoutsCapital

AI built for football

ScoutsCapital runs machine learning across the whole platform, from the moment media is uploaded to the moment a coach sees a better option drawn onto a match clip. Six AI systems are in production today, spanning computer vision, sequence modelling, and football-specific prediction.

Not AI as a feature. AI as the infrastructure.

Computer visionSequence modellingFootball-specific prediction

ScoutsCapital Agent

Digital chatbot interface

Pillar 01

Trust and safety

Monitored around the clock
Audio & video classificationContent moderation models

A safe community, protected by AI

Every piece of user-generated media is screened before it reaches the community. Dedicated models analyse the audio track and the video frames independently, detecting violent content and policy violations. Anything flagged is held back or escalated for human review.

Families, academies, and clubs get a platform that is monitored around the clock, at a scale human moderation alone could not reach.

Screening every upload
  • matchday-warmup.mp4

    Audio + frames analysed

    Cleared
  • u15-highlights.mov

    Audio + frames analysed

    Cleared
  • training-clip-044.mp4

    Escalated to human review

    Held
Pillar 02

Personalisation

Content classificationInterest-profile recommendation

The right content finds you

Posts and their metadata are classified into categories. From there the system builds an interest profile for each user based on genuine engagement, then routes new content to the people most likely to value it.

Scouts see the players and positions they follow. Players see content relevant to their own development. The feed sharpens with use.

Interest profile

StrikersU-17Left wingFinishing

Built from genuine engagement

Match highlight

U-17 striker

Training drill

Finishing

Scout report

Left wing

Sequence modellingRNNsLSTMsTransformers

Advertising that understands intent

Ad targeting runs on sequence models RNNs, LSTMs, and Transformers trained on purchasing behaviour over time. Because these models learn from sequences rather than isolated events, they capture how a user's interests actually evolve instead of over-fitting to a single click.

For advertisers: higher conversion, because ads reach people whose behaviour signals real intent. For users: relevant offers instead of noise.

Behaviour over time, not a single click

Browse
Save
Return
Compare
RNNLSTMTransformer
Real intent, understood
Pillar 03

Football intelligence

Incident classificationCard prediction (yellow / red)

AI that reads the game like a referee

Our card prediction model assesses an incident and predicts whether it results in a yellow card or a red. It's football-native intelligence, the kind of judgement that generic AI platforms have no way to make, and it feeds match analysis, discipline insight, and richer player profiles.

INCIDENT

Incident classification

Yellow card

Reckless challenge

Predicted

Red card

Serious foul play

Feeds match analysis, discipline insight, and player profiles.

Video understandingOn-frame tactical analysis

See the better play

Tactical analysis takes raw match footage and surfaces stronger options directly on the video. Players and coaches don't just review what happened; they see what a better decision looked like, drawn onto the play itself.

Coaching-level insight, generated from footage a player already has.

The pass playedThe better option
Responsible AI

Safety first, by design

Content moderation was the first AI system we built, not the last. Screening media for violence and policy violations before it reaches a young audience is a precondition for everything else on the platform and it shapes how we approach every model that follows.

We publish claims we can stand behind. Where a system is in production, we say so. Where it isn't, we don't.