Data Futurology

Grant Wright delivers genuine substance on Seek's AI architecture (cross-functional team structure, BERT fine-tuning pre-ChatGPT, responsible AI as a strategic function) and the signal-degradation risk from AI-generated CVs is a real insight; however, Jimmy contributes mostly surface-level market observations, and the episode is padded with affirmations, sponsor plugs, and generic 'AI will automate the mundane' commentary.

“we have a jobberta model as we call it, which is fine tuned to our space”

“building about 75 services across 25 countries. Uh, if you include all our services in about five languages”

The 'solved problem' framing for cover letters - questioning whether testing for a skill that AI has commoditised is meaningful - is a genuinely interesting first-principles argument; everything else (human touch in recruitment, GenAI for mundane tasks, AI literacy for the enterprise) is well-worn territory with no contrarian edge.

“if what you're testing for is now a solved problem with this technology, are you really testing for the right thing and is that a skill that's required anymore?”

“I used to recruit for management consulting and the COVID letter was a really great signal. Right? Can someone write very concise cover letter that's kind of a solved problem now”

Grant Wright is a credible practitioner who has built a real AI organisation at scale inside a major global platform, and he speaks with operational authority; Jimmy Acorn is a principal consultant from the show's sponsor (Talent Insights), which introduces a structural conflict and means his contributions function more as client testimonials than expert insight, sharply diluting the overall caliber.

“we started with a couple of people, agreed to a data science team of about 20. Um, now we're in the hundreds with data scientists, product managers, engineers”

“Uh, so I work for a business called Talent Insights, um, we're in Sydney and Victoria. We specialize in recruiting and data and analytics”

Grant supplies concrete numbers - 10 years of AI at Seek, team scaling from ~2 to hundreds, 75 services across 25 countries in 5 languages, and a named fine-tuned model (JobBERTa) - which is substantive; Jimmy provides almost zero specific data, and the host's '7% GDP' claim is uncited and casually dropped without a source.

“building about 75 services across 25 countries. Uh, if you include all our services in about five languages”

“we have a jobberta model as we call it, which is fine tuned to our space”

The host consistently praises rather than probes - stacking affirmations after nearly every response - and at one point explains RAG architecture to guests instead of drawing deeper knowledge from them; the sponsor relationship with one guest forecloses any real pressure, and there is no meaningful pushback or productive disagreement across the full episode.

“That's amazing. Yeah. And um, the acquisitions that Seek has done over the years and um, overseas just I've given you guys such broad, um, data sets and uh, information and where you can apply your AI magic”

“That's awesome. That is great. Um, and then jumping uh, more into the um, LLMs and Engine AI topic”