Leading IT - APAC Insights

The episode covers substantive topics like lambda architecture, Fabric IQ ontologies, and the shift from analytical to operational AI use cases. However, it contains significant filler including lengthy personal anecdotes about visiting Australia, repetitive explanations of the same concepts, and conversational throat-clearing that dilutes the signal-to-noise ratio. A smart operator would extract perhaps 4-5 genuinely novel ideas amid considerable padding.

“data is not enough. We need context and we need to have a curated context, curated knowledge to really power those agents to work effectively”

“if you want to go from analytical to operational, yes, you need to have analytical for. You need to have the analytics for the analytical side, but you need to have the tool to run your operations”

The core distinction between analytical and operational data use cases is useful, and the lambda architecture recommendation (vs. medallion) provides a contrarian angle. However, much of the content recycled common industry narratives: data as foundational to AI, the need for integrated data sources, and the generic benefits of automation. The Fabric IQ ontology concept is the most novel claim but receives insufficient depth to be truly original.

“My belief in architecture is around lambda architecture. So it's not medallion. I think medallion is great for analytics but we want to go beyond analytics”

“if you want to reason about our data, to really not just kind of find the insights because it's easy to find the incidents using deterministic rules or machine learning based rules. But you can use the combination of LLMs and machine learning”

Scott Kesselman is a legitimate operator: Corporate VP leading Microsoft Fabric (the product itself), former 10+ years on Power BI from inception, and prior head of monitoring/observability at Google. He has hands-on product leadership and claims to meet hundreds of customers annually. However, he functions partly as a vendor advocate and the transcript reveals he cannot discuss most future roadmap details, limiting his ability to share truly proprietary operator insights.

“My current role again leading the fabric team within Microsoft. But before that I was with Google leading their monitoring and observability project products. And before that I was with Microsoft for around 10 years on the Power BI team”

“I meet a lot of customers, hundreds, uh, every year, individual companies”

The episode includes two concrete case studies (Apollo Hospitals achieving 40-60% operational efficiency improvement; Veolia moving from GCP to Fabric for cost and real-time intelligence), and mentions specific Fabric products and features (Fabric Real Time Intelligence, Fabric IQ, OneLake, delta format). However, most claims lack supporting data: no metrics on trial F64 capacity, vague timelines ("two or three grand Australian a month" for value with no context), and hand-waved token cost improvements. The bakery example is mentioned only in passing with minimal detail.

“Apollo Hospital, the largest healthcare provider in Asia...they were able to improve their operation efficiency dramatically around 40% all up. I think it was even more than that. I think it was around 60%”

“Veolia which is into energy and clean energy waste management...fully on gcp...they choose fabric, they choose Fabric real time intelligence because it was much, much more cost effective”

The hosts (Josh and Tom/Speaker C) ask substantive follow-up questions and occasionally push back (e.g., "Can I play devil's advocate?", questioning why Power BI alone isn't sufficient, asking about cost justification to CFOs). However, the pushback is gentle and often quickly conceded. The guest frequently launches into long, meandering answers without tight interruption or Socratic probing. Real tension and productive disagreement are absent; it reads as a friendly vendor conversation rather than rigorous interrogation.

“Can I play devil's advocate? You suck...But I have Power BI running on my CRM already. Why do I need, why do I need a data lake?”

“Well, honestly, I've got a lot of customers who keep going back and saying, well, hang on, Data Lake is just a cost”