Data-Driven Podcast

The episode contains a handful of genuinely useful practitioner insights - framing semantic-layer investment as operational risk rather than BI modernization, assigning named business owners to individual metrics, and the observation that AI needs different semantic descriptions than human users. However, large stretches are consumed by host monologuing, mutual congratulation, and repetition of the same 'governed consistent data' refrain.

“Executives don't generally fund infrastructure for infrastructure's sake. They fund risk reduction, scalability and operational efficiency.”

“we're realizing that the descriptions we have on our semantic models for our people users are different than the descriptions we need for our AI users”

The reframe of 'don't sell this as BI modernization, sell it as a control-and-scale problem' is a practically useful and underappreciated distinction. Everything else - self-service always breaking down, AI scaling inconsistencies, single source of truth - is well-worn territory in data circles and offers little a seasoned practitioner hasn't encountered repeatedly.

“we made a point not to position this as a BI modernization effort. We positioned it as a control and scale problem”

“all you're doing is just going to scale inconsistencies and you're going to get wrong answers faster with AI if it's not sitting on top of this foundation”

Both guests are genuine practitioners who have shipped this infrastructure at a real, recognisable enterprise software company - not career conference speakers. Brad brings 20+ years of cross-industry data leadership and Jeremy brings hands-on engineering depth; however, neither is a C-suite decision-maker or widely known operator, and the episode is structured as a vendor case study, which limits candour.

“I've been working in data since 2002”

“I worked with JLL for, ah, about seven years and really focused on the large industrial and tech clients there”

The 30-hour-to-90-second anecdote with an 85% first-pass accuracy claim is the episode's one concrete data point and it lands well, but it is never given numbers, client names, or methodology. Beyond that, the episode is almost entirely abstract - no team sizes, no cost or ROI figures, no named metrics, no timeline milestones beyond '18 months.'

“it was about 85%. And we all kind of had that moment where, like, it's changing”

“dashboards used thousands of times each month”

The host is a vendor executive interviewing his own customers as a thinly veiled case study; he openly calls them 'one of my favorite customers,' answers his own questions at length, and never challenges a single claim. A few process-oriented questions (how did you get executive buy-in, how do you handle business resistance) are reasonable, but there is zero probing, no disagreement, and the interview frequently devolves into the host narrating his own views on semantic layers.

“Well, I can't tell you how happy that makes me feel. That's a, that's like, that's music to my ears.”

“you guys have been really successful and in really getting uh, to a level of maturity very very quickly. Um, so here, uh, one of my favorite customers”