Listen in the Podli app 🎧
Follow your favourite podcasts, listen offline and in the car with CarPlay and Android Auto, and always pick up where you left off. Free to try.
In this week's Data Debrief, Kyle Winterbottom and Catherine Dowden-King unpack a ChatGPT-powered "smart learning" teddy bear aimed at three-year-olds, and use it as a way into a bigger question: when we let technology deliver the output, what happens to the learning journey that used to produce it? That thread runs from toddlers and university degrees all the way into the enterprise.
They also discuss Tuesday's main episode with Chris Pearce, Chief Data Officer at Ageas UK, why AI use cases are finally moving from the sandbox into production, and why the value of that work is still invisible to most customers.
They also discuss:
- Why an AI companion that validates a child's every feeling removes the friction that teaches them how to share, wait and apologise.
- Why "screen-free" is a weak selling point when the device still talks back, listens and adapts.
- How closed-circuit toys like a Toniebox or Yoto player carry a fundamentally different risk profile to a Wi-Fi-connected, always-listening teddy.
- What happens when parental controls protect one side of the conversation but not what the child says.
- Why universities banned AI not to stop augmentation, but to stop replacement — and why that distinction matters everywhere else too.
- How one Strava user overlaid running-route data with rent and income data to find up-and-coming New York neighbourhoods before prices caught up.
- Why personal, intuitive data use cases like that one are a better route into data literacy than heavy-handed formal training.
- Why psychological safety keeps surfacing as the precondition for genuine experimentation with AI.
- How the AI hype cycle has bought data leaders more freedom to test and fail than the analytics era ever did.
- Why podcast guests are suddenly willing to name specific, productionised use cases when a year ago they wouldn't talk on the record.
- What the shift from internally-focused efficiency gains to customer-facing AI means for how organisations talk about their investment.
- Why a business can cut processing times from 100 days to five and still have customers asking what changed for them.
- How the gap between the AI narrative and the actual customer experience is becoming a reputational problem, not just a comms one.
- Why Kyle still had to request a paper form by post to update his details with a pension provider in 2026.
- How Octopus Energy empowering agents to send flowers or waive costs resets customer expectations for every other provider.
- Why data teams need a feedback loop with customers without becoming a ticket office that builds whatever the last complaint asked for.
- What Chris Pearce's point about hallucinations — that nobody ever measured how often tired, stressed humans got it wrong — says about the standard we hold AI to.
- Why the structural and operating model problems inside organisations, not the technology, are what keep use cases stuck in the sandbox.
- How the AI risk conversation has finally given data governance, quality and management their moment of investment.
- Why CDOs should take that funding while it's on the table, whatever vehicle got it there.
