Why OpenAI and Anthropic Are Becoming Consultants

The episode opens with a breaking policy story: the New York Times reported that the Trump administration was considering a formal government vetting process for powerful AI models before public release. This represents a stark reversal from the administration's earlier anti-regulation stance, which had revoked Biden-era mandatory safety testing and emphasized that overregulation could harm U.S. competitiveness against China. The shift appears driven by concerns over Anthropic's unreleased 'Mythos' model and its potential cybersecurity implications, combined with the departure of AI czar David Sachs. Reactions ranged from strong opposition (citing innovation slowdowns and political bias risks) to cautious support from China hawks. Experts noted that effective AI oversight is complicated by poor benchmarks and the difficulty of defining clear risk thresholds, with some arguing that pre-deployment approval is the wrong mental model since models are continuously updated. The story evolved further when it emerged that the U.S. government had already struck early-access agreements with Google, Microsoft, and XAI through the Commerce Department's Center for AI Standards and Innovation.

The main segment focuses on OpenAI and Anthropic both launching large-scale enterprise AI consulting ventures. OpenAI reportedly raised over $4 billion from PE firms to build a 'Deployment Company' valued at $10 billion, while Anthropic officially announced a joint venture backed by Blackstone, Goldman Sachs, and Hellman & Friedman totaling $1.5 billion. Both efforts are modeled on forward-deployed engineering, embedding AI specialists directly within client organizations to build customized solutions. Anthropic specifically targets mid-sized companies and frames engagements as partnerships between its applied AI engineers and clients' internal teams, with a focus on workflow-level transformation rather than just tool deployment.

The host connects these announcements to a broader thesis: there is no AI transformation without organizational transformation. Microsoft's 2025 Work Trend Index is cited extensively, revealing that only 19% of organizations qualify as 'frontier' in both individual AI capability and organizational readiness. A full 50% remain in an 'emergent' state, and the host argues that many of these likely belong in the 'blocked agency' category, where individual AI adoption is high but organizational conditions are lagging. Organizational factors — including culture, manager behavior, and talent practices — account for more than twice the AI impact of individual mindset alone. The episode also highlights three common but ineffective AI adoption shortcuts: 'buy-in hope' (paying for licenses and hoping for transformation), 'contain and delegate' (handing AI goals solely to an AI team), and 'outsourcing of knowledge' (expecting external consultants to figure it out independently). The host notes a potential business model tension for these new consulting ventures, as helping clients be maximally efficient with AI may conflict with the parent companies' goal of maximizing token consumption.