Tuesday, June 2, 2026

6 signals

Perplexity's RevOps leader is replacing hiring decisions with AI agent builds - 3 skills built in 2 months that run autonomously
Voice of Customer automation: Daily-refreshing dashboard using Momentum.io + Salesforce that auto-tags calls, surfaces themes, generates product recommendations, and creates marketing sizzle reels without human intervention
Contrarian thesis: The RevOps scaling playbook is shifting from 'hire specialists' to 'build agents' - credible signal from operator who scaled Ramp and now Perplexity

AI coding agents are rapidly expanding beyond developers - knowledge workers now represent 20% of OpenAI Codex users and growing 3x faster than technical users, with 4M weekly actives (5x growth since February)
The mental cost of AI supervision is emerging as a critical adoption barrier - power users report 'AI psychosis' and cognitive exhaustion from managing multiple fast-moving AI workstreams, creating a new type of workplace stress distinct from traditional productivity fatigue
Agentic AI is creating a workplace artifact integration layer - tools like Codex connect email, calendar, docs, Slack/Teams to surface context across siloed systems, with 60% of users now running concurrent AI tasks (up from <50% in April), signaling shift from single-task automa

AI sticker shock is hitting enterprises hard: 40% of companies see less than 10% cost savings from AI investments, with one CFO accidentally spending $500M/month on Claude
Anthropic's IPO timing is precarious - they've built dominance in enterprise (surpassed OpenAI in business customers) just as their biggest customers are rebelling against AI costs
Even AI executives admit the cost problem is existential: Matt Rogers warns 'risk of enterprises switching to cheaper models is escalating' as open source alternatives match quality without the price tag
The enterprise AI market is entering a reckoning phase where ROI scrutiny will separate sustainable adoption from hype-driven spending - critical inflection point for all AI vendors
Speed of market leadership changes is unprecedented: 'You can't make a three- or five-year bet in this space' - today's leader could be tomorrow's laggard

Microsoft introduced 'average token usage' as a new benchmark metric, signaling industry shift from pure performance to cost-efficiency measurement
Major enterprises (Uber, Salesforce) are hitting AI budget constraints - Uber capped spending after 4 months, Salesforce spending $300M on tokens while freezing engineering hires
New dual benchmark framework: 'intelligence per dollar' replaces tokenmaxxing era - same AI performance at 40% cost difference between providers creates competitive pressure
Entire AI stack must reprice from infrastructure metrics (tokens) to business outcomes (closed tickets, shipped PRs, resolved cases) to align with customer value perception
Era of AI subsidies and unlimited performance-at-any-cost is ending - even most valuable companies cannot afford state-of-the-art intelligence for every use case

Anthropic's Mythos model can detect and exploit security flaws but was withheld from public release
EU initially excluded from access, creating cybersecurity vulnerability concerns in AI-driven threat landscape
Open-source Chinese models expected to reach comparable capabilities within 12 months, democratizing advanced hacking tools