Price:

AI & TECH

AI agent deployment creates enterprise coordination crisis and security gap

Tuesday, June 2, 2026 · from 5 podcasts, 7 episodes
  • AI agents are automating enterprise workflows 100x faster than human processes, collapsing the economic logic of large firms.
  • Legacy security cannot monitor agent intent, treating catastrophic data wipes as legitimate commands.
  • A structural compute shortage is ending subsidized AI, forcing a shift from flat-rate subscriptions to token-based billing.

Coase's law is dead. For nearly a century, firms existed because internal coordination was cheaper than market transactions. Salim Ismail argues on Moonshots that AI has inverted this. The cost of an agent executing a task is now lower than the cost of the meeting to approve it. This 'impedance mismatch' paralyzes legacy organizations while two-person startups deploy agents like Claude to iterate a dozen versions of a project in minutes.

Ismail predicts surviving companies will shrink to 20% of their current staff. They will build AI-native digital twins at the organizational edge, migrating prescriptive workflows like invoice processing to achieve 100x annual performance gains. Middle management faces the steepest cuts as their core function - packaging data for leadership - becomes a native AI capability.

"Companies exist because the transaction costs of doing something inside the firm were lower than doing it in the market. AI has completely flipped that."

- Salim Ismail, Moonshots with Peter Diamandis

This agentic shift is not confined to internal workflows. It is solving fundamental enterprise coordination problems across silos. Pablo Palafox of Happy Robot says logistics is a communication issue, not just a supply chain one. Fragmented data lives in emails, phone calls, and driver conversations where traditional software fails. Voice acts as a 'soft API' for agents to bridge these gaps. Their agents now handle complex, cross-functional workflows for customers like DHL, coordinating across airlines, emails, and phones to track shipments globally.

The explosion of autonomous agents has created a security blind spot. Maxim Bar Kogan of Onyx Security states that legacy tools like identity management lack the context to understand an AI's intent. If an agent is authorized to manage a database, a standard security tool sees a hallucinated 'delete all' command as legitimate. Enterprises grant broad access for productivity, creating a massive surface area for autonomous errors.

"Existing security tools... lack the context to understand the intent of flexible AI agents, creating new control gaps."

- Maxim Bar Kogan, No Priors

This deployment surge is colliding with a physical constraint: a structural compute shortage. Nathaniel Whittemore reports the industry has exited the subsidy era. Power users on $200 monthly plans were consuming up to $10,000 in compute value, a model rendered unsustainable by agentic workflows. Anthropic, GitHub, and Google have all shifted to usage-based billing. Uber burned its entire 2026 AI budget in four months, sparking widespread 'sticker shock.'

The shortage is reshaping the infrastructure landscape. Elon Musk has pivoted, providing access to SpaceX's Colossus data centers to ease Anthropic's compute constraints. This turns SpaceX into a neocloud provider ahead of its IPO. The economic unit of AI has officially moved from the person to the token.

Companies capturing value are those building institutional coordination layers, not just deploying individual tools. Data cited by Whittemore shows 20% of companies capture 75% of AI's economic gains. At Ramp, the internal AI platform 'Glass' auto-configures for every employee, providing 350 pre-built skills to ensure individual breakthroughs become company-wide baselines. Without such a harness, George Zarkadakis argues, individual AI creates faster organizational chaos, not collective intelligence.

The transition is a race against internal resistance and external vulnerability. Ismail notes that 44% of Gen Z workers sabotage AI training to protect their jobs. Meanwhile, any high-margin business line can be replicated by a small team using agentic tools in 60-90 days. The firms that survive will be those that stop trying to fix the mothership and instead build a parallel, AI-native entity at its edge.

Source Intelligence

- Deep dive into what was said in the episodes

The AI Token Shortage Begins [AI Monthly Recap]Jun 1

  • Nathaniel Whittemore argues the AI industry is experiencing its second major transitional moment of 2026, moving from an AI subsidy era to a token scarcity era defined by structural compute shortages driving up costs.
  • Foundation model company revenue shifted from seat-based subscriptions to API token consumption as the primary economic unit, fueling explosive growth. OpenAI surged to $30B ARR and Anthropic reached a $47B annualized run rate by mid-2026.
  • Uber burned through its entire 2026 AI budget in four months, and its COO later expressed skepticism about the value derived, sparking a broader 'AI sticker shock' conversation in corporate America.
  • In response to unsustainable costs, providers are shifting from flat-rate subscriptions to usage-based billing. GitHub Copilot, Google Gemini, and Anthropic all announced changes, with Anthropic's move to per-token billing for third-party tools causing significant user backlash.
  • Nathaniel Whittemore contends companies face a massive 'capabilities overhang' where agentic AI potential far outstrips organizational ability to adopt it, prompting OpenAI and Anthropic to launch major enterprise consulting initiatives.
  • AI infrastructure is 'going vertical'. Inference provider Base10 is raising $1B at an $11B valuation, and OpenRouter raised a $13M Series B, as companies build solutions for the costly, constrained compute environment.
  • Elon Musk pivoted from promoting Grok to partnering with Anthropic, providing access to SpaceX's Colossus 1 and Colossus 2 data centers to ease Anthropic's compute constraints, effectively turning SpaceX into a neocloud provider ahead of its IPO.
  • Wall Street is favoring AI infrastructure stocks, with memory companies like SK Hynix and Micron reaching trillion-dollar valuations. Meta is also considering becoming a cloud business to monetize its massive compute investments.
Also from this episode: (5)

AI & Tech (5)

  • The token shortage is driving market-based innovation for cheaper inference. Cursor's Composer 2.5 offers lower cost than top models, while DeepSeek made a permanent 75% price cut on its V4 model to capture cost-conscious users.
  • Model releases are becoming incremental, shifting focus to the harnesses and applications. Riley Brown and Greg Eisenberg both noted that updates to environments like Claude Code and Codex now matter more than modest model improvements.
  • Sam Altman and Dario Amodei softened their public narratives about AI's disruptive impact on jobs, with Altman citing new evidence he overestimated the transformation's speed, opening space for more nuanced policy discussions.
  • AI policy debates are fracturing, with some Democrats like Bernie Sanders and AOC calling for data center moratoriums, while Elizabeth Warren advocates for novel taxation structures like token taxes instead of blocking development.
  • The White House involved itself in model release governance, partly opposing wider access to Anthropic's Mythos model due to concerns over the token shortage and a desire to reserve compute for government use.

How to Use /Goal to Do More With AIMay 31

  • Nathaniel Whittemore cites a PwC study showing 75% of AI's economic gains are captured by only 20% of companies, marking a widening performance gap.
  • Whittemore reports AI-leading companies are 2-3x more likely to use AI for growth opportunities and 2.6x more likely to report AI-enabled business model reinvention, according to PwC.
  • McKinsey's AI transformation manifesto identified 12 themes separating leaders from laggards, arguing AI leaders focus on economic leverage points, not just efficiency.
  • McKinsey found AI transformations in leading companies delivered a 20% EBITDA uplift, broke even in 1-2 years, and generated $3 incremental EBITDA for every $1 invested.
  • George Zarkadakis argues institutional AI requires coordination layers to align individual AI use, as uncoordinated AI adoption creates organizational chaos.
  • McKinsey contends over 70% of AI talent should be in-house, as AI transformation is fundamentally a people transformation, not just a tech implementation.
  • Ramp built its internal AI workspace Glass to solve coordination problems, auto-configuring connections to all company tools via SSO so agents operate with full organizational context.
  • Seb Go argues the biggest failure mode in AI adoption is isolation, where one employee's discovered workflow doesn't help others, which Ramp solves with its Dojo marketplace for reusable agent skills.
  • Ramp's Dojo marketplace contains over 350 reusable AI skills, with an AI guide called Sensei surfacing the most relevant five skills for a new employee based on role and tools.
  • Seb Go states Ramp built Glass in-house because internal productivity is a moat, ownership allows same-day fixes, and solving internal problems directly informs their external AI products.
  • Whittemore notes OpenAI reported 50% of Codex usage is not about coding, highlighting AI's broader utility for knowledge work beyond software engineering.
Also from this episode: (2)

AI & Tech (2)

  • Ryan Carson predicts complete end-to-end code factory solutions from major AI players by year-end, eliminating the need to duct-tape disparate tools together.
  • Vercel CEO Guillermo Rauch announced the company is open-sourcing a reference platform for cloud coding agents, following patterns used by Stripe, Ramp, Spotify, and Block.

The Annual AI Slowdown Panic is HereMay 27

  • Whittemore argues the 'real-world recalibration' phase emerges from physical and economic constraints, like the structural compute shortage, which forces usage-based AI pricing and curbs limitless experimentation.
  • Anthropic shifted enterprise customers from a subsidized $200 flat rate to usage-based pricing at $20 per seat plus tokens. GitHub Copilot user screenshots show hypothetical bills rising from tens or hundreds of dollars to thousands under the new model.
  • Policy proposals are emerging from a recalibrated perspective, like Matthew Yglesias's idea to mandate affordable compute set-asides in data centers or Mark Cuban's call for a small federal tax on AI tokens to fund debt reduction.
Also from this episode: (8)

AI & Tech (7)

  • Nathaniel Whittemore introduces the 'AI doom cycle,' a five-stage emotional framework for public engagement with AI: Skepticism/disbelief, AI psychosis, doom desperation, real-world recalibration, and enlightened excitment.
  • Ken Griffin reversed his January 2024 AI skepticism, citing new coding models. Griffin says AI now completes high-level finance research in hours versus human weeks, delivering a 15-25% productivity boost at Citadel.
  • Media re-surfaced predictions from AI CEOs like Mustafa Suleyman and Dario Amodei. Suleyman gave an 18-month timeline for AI automating all white-collar work, while Amodei predicted 10% overall and 50% entry-level white-collar unemployment.
  • Graduates at University of Arizona and University of Central Florida loudly booed commencement speakers who discussed AI's role in the future, a reaction tied to AI leaders' own job displacement warnings.
  • Meta is laying off about 8,000 employees, roughly 10% of its staff, as part of an AI restructuring, with low morale exacerbated by screen-tracking software used for AI training.
  • Whittemore's 'enlightened excitment' state enables nuanced discourse, like economist Alex Emos's essay on a rising 'relational sector' or Jensen Huang's Carnegie Mellon speech focusing on re-industrialization and new job creation.
  • Sam Altman recently shifted OpenAI's public narrative from doom to augmentation, stating the goal is to build tools that 'augment and elevate people' and sharing personal anecdotes about AI improving work-life balance.

Startups (1)

  • Didi Desai's viral post describes a bifurcated Silicon Valley where ~10,000 employees at top AI firms gained over $20M in wealth, fueling widespread career anxiety, malaise, and layoff fears among those outside that group.

Building AI Agents for Enterprise OperationsJun 1

  • Pablo Palafox says Happy Robot started by solving enterprise coordination problems in logistics, not just supply chain issues, after seeing a co-founder manually track olive oil shipments.
  • Luis Parag explains the company's early focus was the technical challenge of building a realistic voice agent for phone negotiations, which required fine-tuning models like Mistral and Llama for speed and building custom agent infrastructure.
  • Happy Robot's current enterprise customer base includes nine of the top ten US freight brokers, seven of the top ten tracking companies, and two of the largest ocean carriers.
  • Luis Parag argues the biggest problem for voice AI is conversation handling, not latency or realism. Agents must know when to talk, handle interruptions, filter background noise, and reason asynchronously.
  • Pablo Palafox states their agents now handle complex, cross-functional workflows like freight forwarding customer support, which requires coordinating actions across airlines, emails, and phone calls to track shipments.
  • Luis Parag says their platform uses a deterministic approach for tasks like rate negotiation, where an agent requests permission from an external tool instead of being exposed to maximum rates, to prevent hallucination.
  • The company runs collections campaigns for a large supply chain customer, making 20,000 to 50,000 daily outreach calls to recover duties on parcels.
  • Pablo Palafox describes a 'pyramid of work' where simple, high-volume tasks form the base, but the real economic value is in complex, strategic decisions at the top, which require shared context across functions.
  • Luis Parag argues that deploying agents to execute work is the best way to clean and connect enterprise data, as agents consistently populate systems, unlike humans who forget or make errors.
  • Happy Robot uses a forward-deployed engineering team to scope and build solutions, which Pablo Palafox says acts as an extension of the product team to capture customer context and accelerate implementation.
  • The company's work with DHL involves over 40 agents deployed across 80 countries, sharing context across regions and functions, which revealed they were solving a general enterprise coordination problem.
  • Pablo Palafox says they are now being pulled into telecommunications, utilities, and insurance because these industries face similar coordination problems between customers, partners, and employees.
  • Luis Parag states they focus on problems where communication is the interface to the external world, like voice, email, or web browsing, and where standard operating procedures are unclear or highly contextual.
  • Pablo Palafox emphasizes the importance of a human-like experience, noting that even when disclosed, end-users quickly forget they are talking to an AI if the conversation flows naturally and helpfully.

Building an AI Guardian for Enterprise with Onyx Security CEO Maxim Bar KoganMay 28

  • Onyx Security trains models and builds agents to oversee other AI agents, aiming to detect illegitimate actions as enterprise AI adoption grows exponentially.
  • Maxim Bar Kogan states autonomous agents like coding assistants are the fastest-growing category in enterprises, outpacing low-code automation platforms.
  • Kogan argues existing security tools like identity management and API monitoring lack the context to understand the intent of flexible AI agents, creating new control gaps.
  • Onyx's approach uses small, specialized models to efficiently flag high-risk AI actions, reserving more powerful analysis only for critical moments to manage cost and latency.
  • Kogan sees Mithril-level AI models dramatically lowering the cost of vulnerability discovery, forcing security teams to implement foundational controls quickly.
  • He believes independent oversight is crucial because enterprises distrust vendors auditing their own products and need solutions adaptable to multiple AI providers.
  • Kogan notes enterprises are unwilling to share historical agent behavior data with Anthropic or OpenAI due to those companies' data-hungry training practices.
  • Auto-GPT's early demonstration of autonomous agents ignited market imagination, highlighting the potential and risks of AI performing complex computer tasks.
  • Kogan asserts the core challenge of AI oversight is interpreting agent intent, not just proxying data, which requires understanding what AI systems 'think'.
  • Onyx's founding insight was the need to control increasingly smart AI agents, especially as they begin managing critical infrastructure like power grids.
Also from this episode: (4)

Startups (1)

  • The Israeli tech ecosystem excels at understanding security team workflows and building products tailored to their daily operational needs, according to Kogan.

AI & Tech (3)

  • He predicts mechanistic interpretability of AI models will advance significantly as smarter AI systems emerge, aiding in understanding and controlling intelligence.
  • Financial institutions like JP Morgan adopt AI cautiously due to high risk profiles, contrasting with startups that aggressively deploy agents to gain competitive edge.
  • Kogan advocates for gradual, controlled release of Mithril-level AI models to allow enterprise security teams time to develop defenses and prevent catastrophic failures.

NEAR’s AI Money Thesis: Intents, Privacy, and Tokenomics | Sal TernulloMay 27

  • Ternullo cites cloud infrastructure adoption in regulated finance as an analogy: enterprises resisted shared services until security and compliance controls matured, a cycle now repeating with private, user-owned AI.
  • Sal Ternullo's company, Sovereign, is a NASDAQ-listed Near treasury firm acting as a commercialization partner. It drives adoption by participating in governance and onboarding MPC node operators.
Also from this episode: (7)

AI & Tech (6)

  • Sal Ternullo describes Near as AI money, a settlement layer for agents to transact on behalf of users in the real economy. Intents and Iron Claw are the technical embodiment of this vision.
  • Near Intents has processed almost 20 billion in total volume, generating over 30 million in fees. Its switch in February now directs all fees to buy back and burn the Near token.
  • Zashi, Venice AI, and Infinex are prominent third-party integrations using Near's technology. Zashi alone has generated over 3 million in fees from cross-chain swaps and Zcash access.
  • Ternullo argues Near is currently undervalued based on human-driven Intents usage, but offers a 20-50x opportunity if billions of AI agents adopt its settlement and privacy infrastructure.
  • Ternullo sees Layer Zero as Near Intents' primary competitor in chain abstraction, but argues Near offers different design tradeoffs and has captured significant market share.
  • Near enables confidential transactions through NearCON, offering a retail-friendly self-custody solution with privacy toggles for every transaction.

Protocol (1)

  • Near's tokenomics changed in October, reducing protocol emissions to 2.5% annualized inflation. Value accrual to the token is now driven by vertically integrated products like Intents, not base-layer fees.

The Organizational Singularity: AI-Proof Your Company | EP #258May 26

  • Ismail says companies must architect around intelligence instead of hierarchy, moving from human-centric workflows to AI-native, agentic systems.
  • Ismail warns any high-margin business line can be replicated by a small team using tools like OpenClaude in 60-90 days, making incumbent firms vulnerable.
  • Companies should build an AI-native digital twin at the edge, migrating workflows like invoice processing to it while leaving the legacy core untouched.
  • Ismail estimates a successful transition yields 100x performance improvements per year, and companies can eventually operate with 10-25% of their current workforce.
  • Middle management will shrink by 60% as coordination tasks vanish, while the C-suite shifts to dashboard oversight and exception handling.
  • Ismail's intelligence stack has six layers: purpose, sensing, interpretation, decision, orchestration, and learning, wrapped by a governance protocol for agent oversight.
  • Agent passports with metadata constraints prevent rogue actions, supported by granular rollback, searchable logs, and human review queues.
  • Ismail cites Cognitions Labs achieving 73x ARR growth after going fully AI-native, and says the full industry transition will take five to seven years.
  • Ismail's rewrite methodology involves backcasting from a future vision, scoring the company on seven dimensions like organizational drag and AI integration.
Also from this episode: (5)

AI & Tech (4)

  • Salim Ismail argues Coase's 1937 'The Nature of the Firm' model is obsolete because coordination costs inside a company now exceed external execution costs due to cheap AI agents.
  • The 'organizational singularity' centers on recursive self-improvement at the workflow level, enabling companies to learn faster than competitors.
  • Ismail notes 44% of Gen Z workers sabotage AI training to protect their jobs, exemplifying the organizational immune system that blocks change.
  • The new book 'Organizational Singularity' will be released as a Claude AI skill to stay updated, not as a static text.

Health (1)

  • Peter Diamandis highlights Fountain Life's full-body MRI and early cancer detection screening, noting 3.3% of members had undiagnosed cancers.