Reported execution: Serena versus Serena
AKQA describes Never Done Evolving as Nike's 50th-anniversary story about Serena Williams' development rather than a retirement retrospective. The team analysed official tournament data and archival footage, adapted Stanford's vid2player technique, and modelled the playing styles associated with 1999 and 2017. The campaign turned those models into a virtual match, livestreamed with commentary and extended through media placements.
The case is specific about the useful unit of AI work: it did not claim a model had created athletic greatness. It used data and video analysis to simulate a contest between two bounded representations of one athlete's game.
Our reading: let the technical device serve a question
The campaign works because its technical premise can be stated as a sporting question: how did one player change? A virtual matchup gives an audience an intelligible way to encounter an otherwise invisible progression in movement, tactics and experience.
That is a stronger starting point than treating synthetic sport as spectacle by itself. For a comparable piece, identify the question a fan already has, then decide whether simulation can answer it more clearly than archive footage, a graph or an interview.
Do not confuse a simulation with a historical match
AKQA says the system generated many simulated games and matches. Those outputs are model results, not recovered facts about an event that happened. Editorial framing should keep that distinction visible, especially when simulation is presented with the conventions of broadcast sport.
Teams should also record the dataset boundary and the interpretive decisions needed to turn footage into playing behaviour. This preserves the human judgement in the method and gives critics a way to question the result without dismissing the work as a trick.
Evidence boundary: the agency's result is not a sports finding
AKQA reports view and impression outcomes, but the public account does not supply a comparison design that would isolate the simulated match from Serena Williams' audience, Nike's anniversary media or the live-event format.
The defensible lesson is narrower: machine learning made a career-evolution story newly playable. It does not prove that simulation is automatically the most persuasive way to tell an athlete's story.
Sources & limits
Agency case study; audience and performance results are reported by AKQA, not independently audited.
- AKQA — Nike Never Done Evolving — source publication date not stated.
Send a correction with the passage and supporting source.