How to Decide What Work AI Should Do for You: The AI Deputization Audit
Summary
This episode analyzes the shift from AI capability to context, highlighting GrokBot’s “Teach a Task” and ChatGPT’s “Computer History” features. It reviews Google’s fast but mid-tier Gemini 3.7 Flash and OpenAI’s Ultra Fast Mode, alongside a study showing US models often outperform cheaper Chinese alternatives in efficiency. The host introduces an “AI Deputization Audit” to help listeners identify which frequent, low-stakes tasks are best suited for automation. Finally, it covers OpenAI executive turnover and emerging user strategies for deploying AI agents.
Topics discussed
Intro: AI automation math and new context tools
Google Gemini 3.7 Flash: Speed, cost, and benchmarks
AlphaSense study: US vs Chinese model efficiency
OpenAI Ultra Fast Mode and executive turnover
Sponsors: KPMG, Blitzy, Harbor, and HyperAgent
GrokBot Teach a Task vs ChatGPT Computer History
The AI Deputization Audit framework and scoring
Identifying blockers and GrokBot user feedback
Listen ad-free on Castria