Julie Souza spent six years as Global Head of Sports at AWS, so if you worked in sports technology between 2020 and 2026 you probably met her in a pod at IBC or NAB, and she is now Executive in Residence at PwC. We crossed paths at many of those events and I always admired the energy, and that on stage she brought her whole self, humour and opinions included, which is rare for someone carrying a badge that size. I had her on A Guy With A Scarf this week, and I want to start with the test she ran on her own team.***"If we can't explain it in three big buckets, then there's not a story there."*** The buckets are the business problem, the solution deployed, and the outcome it drove with a number attached, and her warning for 2026 is that "we want an AI strategy" is landing on her desk in exactly the shape that "we want a locker room in the metaverse" landed four years ago.That metaverse brief came from a professional team, name withheld, and her first question was ***"why do you want to do that?"***, the answer was that it was the metaverse, so she asked about customer acquisition and the revenue stream, ***"and they sort of just blinked at me."*** Work backwards, she says, from what you are trying to do better, cheaper, faster or at greater scale, and the AI briefs she sees mostly arrive without any of those four words in them.I challenged her on the test, because the business lines I built that ran for a decade were built when nobody was measuring me. The horizon has to be long, she said: ***"If you're looking for quick wins, you might have a few of those, but I don't think then you're using it to its full power either."*** When the NFL changed its kickoff rule, the league and AWS ran the proposal through 10,000 simulated seasons to find a play that was less injurious and more exciting at once, and the AP reported in October 2025, using the NFL's own figures, that concussions fell 17% from 2023 to 2024. The proof on a programme like that arrives seasons in, so a quarterly review would have killed it in year one.On who should hold a multi-vendor solution together, her word is quarterback, ***"somebody you trust quarterbacking that"***, and on doing it all in-house: ***"If anybody's telling you they can do it all, that's probably not true."***On AI more widely she named two things: data centres becoming politically fraught in the US, and reports of frontier models jumping their guardrails, with no regulation in sight. To the argument that rules mean falling behind in the global race: ***"What are we racing toward? Like human extinction? Because I'm fine falling behind on that."*** On jobs, the usual line is that entry-level work goes and higher-judgement work remains, and her counter is that ***"how you know to make high judgment decisions is through experience."*** Stanford's "Canaries in the Coal Mine" paper, revised in August 2026 on ADP payroll data, finds 22 to 25 year olds in the most AI-exposed jobs 19% below where they would be against less-exposed peers, and no gap for experienced workers. I pushed back with the Google and Wikipedia panics, and with my laptop running a Claude Code project all day, and she replied that with Google you still wrote up your own analysis, while agentic AI does the writing up too, and for her the writing up is the thinking. Then she pointed out that I am building on decades of rigour and an eleven-year-old is not.She closed on live sport: ***"I don't think live sports can be disintermediated."*** Watch the video, and if you are working out what AI is for, write to me with the problem and the number you want to move, and I will tell you if it passes her test.Sources- AP via NBC Washington, 14 October 2025: https://www.nbcwashington.com/news/sports/nfl/nfl-ai-predict-injuries-keep-players-healthier/4001576/- Stanford Digital Economy Lab, Canaries in the Coal Mine (rev. August 2026): https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine