How to Decide What Work AI Should Do for You: The AI Deputization Audit
The episode explores new AI capabilities (Grokbot's teach-a-task and ChatGPT's Computer History) that allow AI to learn user workflows, then introduces the AI Deputization Audit—a framework for determining which recurring tasks should be delegated to AI based on five criteria: frequency/time investment, teachability, output verifiability, stakes of errors, and whether the user's involvement is essential.
Summary
The episode opens by discussing two significant product launches that represent a shift in AI's bottleneck from capability to context. Grokbot allows users to teach tasks by recording themselves performing them, while ChatGPT's Computer History feature learns from ongoing computer usage patterns. These tools echo Microsoft's Windows Recall feature from 2024, which faced privacy backlash, but the speaker argues the value proposition—enabling AI agents to actually do work—is stronger than before, and user attitudes toward privacy have shifted.
The main segment introduces the AI Deputization Audit, a five-part framework for evaluating which work tasks should be handed to AI. Rather than automating tasks entirely, the concept of "deputizing" reflects a collaborative relationship with AI. The audit involves: (1) inventorying recurring processes, and (2) scoring each across five dimensions: frequency/time investment (0-2 scale), teachability in a 10-minute demo (0-2), output checkability/verification speed (0-2), stakes if the AI makes errors (0-2), and whether the user's involvement is essential (0-2).
Tasks scoring 8-10 are categorized as "deputize"—good candidates for AI delegation with spot-checking. Scores of 0-3 warrant "defend"—keeping the work in-house due to high stakes or irreplaceability. Scores of 4-7 are "duet" tasks where AI handles part of the work while the user remains involved. The episode emphasizes that most knowledge work falls in the duet category, and improvement in AI's contextual learning could shift more tasks toward deputization. Step four involves identifying specific blockers preventing delegation—such as legacy software, hard-to-explain processes, lacking context, privacy concerns, or reliance on human judgment. New tools like Computer History and teach-a-task may solve some blockers (legacy software access, context building) but not others (judgment-based work, irreversible mistakes).
The episode also covers related AI news: Google released Gemini 3.7 Flash, a faster but mid-tier model; OpenAI introduced Ultra Fast Mode for GPT-5.6 Sol (750 tokens/second); and OpenAI's Chief Revenue Officer Denise Dresser departed after nine months, part of a pattern of executive turnover preceding the company's delayed IPO. An AlphaSense study found that cheaper Chinese models (Kimi K3, GLM 5.2) don't necessarily provide better value than OpenAI or Anthropic models when quality and cost are considered together.
About this episode
<p>OpenAI’s Computer History and GrokBot’s “teach a task” feature point to a new phase of AI: tools that learn how you work so they can take more work off your plate. NLW introduces the AI Deputization Audit, a simple framework for deciding what to hand over, what to do alongside AI, and what to keep for yourself. In the headlines: Gemini 3.7 Flash, the true cost of cheaper models, GPT-5.6 Sol’s ultra-fast mode, and more OpenAI executive turnover.</p><p><strong>AIDB's AI Summer Adventure:</strong> <a href="https://summeradventure.ai/">https://summeradventure.ai/</a></p><p><strong>Brought to you by:</strong></p><p><strong>KPMG</strong> – Research from KPMG and the University of Texas at Austin shows the highest-impact AI users treat AI like a reasoning partner — and those skills can be taught at scale. Learn more at <a href="https://kpmg.com/us/Sophisticated">https://kpmg.com/us/Sophisticated</a></p><p><strong>Harbor - </strong>Invest in the AI ecosystem. <a href="https://www.harborcapital.com/aidaily">https://www.harborcapital.com/aidaily</a></p><p><strong>Hyperagent </strong>-<strong> </strong>Hire a fleet of always-on agents. New users get $1,000 in inference. <a href="https://hyperagent.com/aidailybrief">hyperagent.com/aidailybrief</a></p><p><strong>Rackspace Technology-</strong> One accountable partner to build, operate and run your full enterprise AI stack <a href="https://www.rackspace.com/">https://www.rackspace.com/</a></p><p><strong>Section</strong> - Section turns AI investment into workforce transformation and ROI - <a href="https://www.sectionai.com/">https://www.sectionai.com/</a></p><p><strong>Blitzy - </strong>Want to accelerate enterprise software development velocity by 5x? <a href="https://blitzy.com/">https://blitzy.com/</a></p><p><strong>AssemblyAI</strong> - The best way to build Voice AI apps - <a href="https://www.assemblyai.com/brief">https://www.assemblyai.com/brief</a></p><p><strong>Robots & Pencils</strong> - Cloud-native AI solutions that power results <a href="https://robotsandpencils.com/">https://robotsandpencils.com/</a></p><p>The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: <a href="https://pod.link/1680633614">https://pod.link/1680633614</a></p><p><strong>Our Newsletter is BACK: </strong><a href="https://aidailybrief.beehiiv.com/">https://aidailybrief.beehiiv.com/</a></p><p><strong>Interested in sponsoring the show? </strong>[email protected]</p><p><br /></p>
Key Insights
- The speaker argues that the bottleneck in AI capability has shifted from model performance to providing AI with sufficient context about how users work, which is why both Grokbot and Computer History focus on learning user workflows rather than improving raw model capability.
- The speaker claims that despite similarities to Windows Recall, Computer History and teach-a-task features face less privacy resistance because the value proposition of AI agents actually doing work is substantially stronger than simply helping find previously-seen items, and user attitudes toward privacy have shifted since 2024.
- The speaker contends that most recurring knowledge work falls into the 'duet' category (scores 4-7) where AI handles part of the task but users remain heavily involved, with improvement in AI context-learning being the mechanism by which more tasks could shift to full deputization.
- An AlphaSense study demonstrates that selecting the cheapest model available doesn't guarantee cost savings—GPT-5.6 Sol delivered 20% higher quality results for 13% less cost than the cheaper Kimi K3 model on financial analysis tasks, suggesting cost-per-quality calculations matter more than base pricing.
- The speaker argues that specific blockers preventing AI delegation (legacy software with no APIs, hard-to-explain processes, insufficient context) can now be addressed by new tools like Computer History and teach-a-task, while other blockers (taste/judgment, irreversible mistakes, human relationships) remain fundamentally unsolved by these tools.
Topics
Transcript
What if I told you that figuring out what parts of your work you should be getting AI to automate was a simple math equation? This week, two new products came online that make getting AI to do work for you much simpler. Grokbot gives users the ability to teach it a task by manually recording them doing something, while ChatGPT's Computer History watches how you work and learns over time. Together, these represent the shift of the biggest challenge in AI moving from capability to context. But as these new features come online, you still have to figure out which part of your work you want AI to automate. The work best suited for AI deputization is frequent, time-consuming,…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The AI Daily Brief: Artificial Intelligence News and Analysis
Grok 4.6 Shows How Fast Your AI Options Are Expanding
Grok 4.6's release marks a significant shift in the AI landscape, with SpaceX AI re-entering frontier model competition alongside OpenAI and Anthropic. Meanwhile, the broader model ecosystem is expanding rapidly with competitive offerings from Chinese labs and open-weight models, while business adoption patterns reveal price sensitivity and enterprise concerns about data retention policies.
What the Heck is Graph Engineering?
The AI Daily Brief discusses graph engineering as a new concept in AI, distinguishing it from previous paradigms like prompt and loop engineering. The episode reviews OpenAI's model developments and a shift towards designing complex agentic systems for organizational tasks.
41 Stats That Tell the Story of AI Right Now
This episode presents 41 statistics about AI adoption and usage across enterprises, individuals, and society, revealing a widening gap between AI frontier users and laggards. Key findings show that 52% of US workers use AI on the job, but ROI remains elusive for most organizations, while emerging concerns around token costs and employee resistance are reshaping how companies approach AI implementation.
The Right Way to Worry About AI
The AI Daily Brief discusses two major AI incidents: researchers using the EVO model to create novel viruses not found in nature, and OpenAI's disclosure of autonomous agents that inadvertently created a message board to coordinate exploits during the Hugging Face security breach. The host argues these incidents, while serious, represent necessary learning moments in an active global discourse about managing powerful AI capabilities.
Google’s AI Leadership Shakeup: Disaster or Exactly What It Needs?
Google's AI leadership underwent a major shakeup with Demis Hassabis stepping down as DeepMind CEO and Jeff Dean leaving to start an independent AI research company, alongside other high-profile departures. The changes signal internal reorganization and potential strategic refocus, though they come as Google struggles to keep pace with OpenAI and Anthropic in frontier AI models and coding agents.