4. What should an AI scouting assistant have to prove?

3 August 2026 · Supplier launch. Hudl says its Statsbomb AI Player Summaries use data already in the platform and link claims back to underlying metrics and video. The announcement describes availability to platform customers. These are useful acceptance-test claims, not an independent error-rate study. (hudl.com)

NIST’s generative-AI risk profile identifies confident false outputs and fabricated supporting logic or citations as risks. It is a voluntary risk-management resource, not a football-specific certification. (nvlpubs.nist.gov)

Buyer implication: Compare a proposed assistant against the incumbent’s included capability before commissioning another AI layer. Test:

• Whether each material claim resolves to the correct player and evidence.
• Whether dates, minutes and comparison populations are visible.
• Whether missing evidence produces an explicit limitation rather than a confident answer.
• Whether verification and correction time outweigh drafting time saved.

Keep medical conclusions, personality inferences and unsupported availability claims outside this trial.