In this episode of the Data Debrief, Catherine Dowden-King and Kyle Winterbottom reflect on the week's news and the latest Driven by Data: The Podcast interview, with Tom Sadler of HP. Their central question is whether organisations are measuring AI activity (headcount, spend, job titles, tooling) instead of the productivity and financial outcomes that matter.

They begin with the ASOS breach and the questions it raises about vendor risk. They then work through the latest Evident AI Index and what it rewards, before turning to Tom's candid views on hardware, cost and culture.

They also discuss:

• Why the ASOS ransom note, delivered through the company's own app, marks a new and uncomfortable form of attack.
• How data leaders should work in lockstep with the CISO, or with the CIO where there is no CISO, as vendor ecosystems grow.
• Why the full story of a breach like this can take months to emerge, and why speculation helps no one.
• Why only 12% of reported AI use cases in the Evident AI Index demonstrate operational impact, and just 1% report financial impact.
• How the index's weightings (45% talent, 30% innovation, 10% transparency) shape what it rewards.
• Why organisations with the strongest financial returns from AI may not appear on the index at all.
• What the "too early to judge" argument gets right, and why 1,100 use cases over five years still demands better measurement.
• Why leaders chasing a higher ranking risk spending more money for the same 1% return.
• How adding "AI" to a job title, without mandate or budget, repeats the chief data officer pattern.
• Why the market is starting to reward titles over delivery when it hires.
• Why Tom Sadler argues hardware should be the last thing you think about in an AI strategy.
• How local compute can cut experimentation costs compared with spiralling cloud token costs.
• What running AI locally means for environmental impact.
• Why the CDO and CIO sometimes have never met, and how that drives sprawling costs and shadow AI.
• How a confined local "sandbox" lets teams experiment safely, with successful ideas scaled across the business.
• Why culture, not tooling, decides whether you can spot what has worked, and why a use case that cuts five hours of work to 30 minutes needs a place to be shared.
• What it takes to build the honesty to review mistakes regularly and correct them.
• Why a chief financial officer's advice to start running in your old trainers applies to AI spend.