FEATURED

Gartner Predicts 40% of Enterprises Will Scrap Their AI Agents — Here’s How to Beat the Odds

Gartner dropped a sobering forecast this week: by 2027, 40% of enterprises that deployed autonomous AI agents will demote or decommission them — not because the tech failed, but because governance gaps were discovered only after incidents occurred.

ZDNet spoke with data leaders at Whoop, Fanatics, and Synopsys who are beating the trend. Their playbook has three legs:

Why this matters for you: The 40% failure rate isn’t about bad AI. It’s about bad deployment. Companies that treat agents as plug-and-play will join the scrap heap. Companies that build governance, context, and human coaching into their agent workflow will capture the gains.

🔗 ZDNet — Full story

QUICK HIT

The “Judgment Tax”: Why Brittle RPA Is Bleeding Your Team Dry

Forbes Tech Council published a sharp analysis this week on what they call the “Judgment Tax” — the hidden cost of having skilled operators babysit brittle automation. The numbers are brutal: 45% of companies experience RPA bot failures at least once a week, and 79% need advanced programming skills just to maintain their bots.

The alternative? UI-layer AI agents that handle unstructured judgment work — not just clean, rules-based tasks. One mid-sized mortgage bank used this approach to collapse a 3-5 day manual underwriting form process into a few hours, with operations staff tuning the agent directly (no engineers needed).

🔗 Forbes — Full story

QUICK HIT

Google Bets $50M on Welders and Electricians for the AI Buildout

In a counterintuitive move, Google.org is investing $50 million to train 300,000 American workers — not in prompt engineering or ML ops, but in welding, electrical work, plumbing, and fiber tech. The initiative partners with 14 labor unions and four trade associations.

Sundar Pichai’s logic: America’s AI infrastructure — data centers, fiber networks, power systems — needs a physical workforce that doesn’t exist yet. The $1.8 trillion AI infrastructure boom is bottlenecked not by chips, but by the skilled tradespeople who install them.

🔗 MSN / Google — Full story

US Blocks Foreign Access to Anthropic’s Most Powerful Models

Anthropic has taken its latest frontier models — Fable 5 and Mythos 5 — offline for all foreign nationals, complying with a Trump administration export control directive issued under the June 2 executive order on AI national security. The move is the most aggressive US government intervention in AI access to date.

Anthropic President Daniela Amodei pushed back publicly, calling the government’s handling a “misunderstanding” that lacks transparency and grounding in “technical facts.” The company says it supports government safety review — but through a statutory process, not ad-hoc directives.

The ripple effects are immediate: India’s tech community erupted, with Zoho’s Sridhar Vembu declaring “globalisation is dead” and Mohandas Pai calling for a sovereign AI mission. The export controls are reshaping the global AI landscape in real time.

🔗 Daily Camera — Full story | Deccan Herald — India reaction

G7 Summons Altman, Amodei, and Hassabis for Historic “Algorithmic Diplomacy” Session

For the first time, the CEOs of all three major Western AI labs — Sam Altman (OpenAI), Dario Amodei (Anthropic), and Demis Hassabis (Google DeepMind) — will attend a G7 summit together. The meeting runs June 15-17 in Évian-les-Bains, France, with AI placed prominently on the agenda by French President Macron.

The timing is electric. Both OpenAI and Anthropic have confidentially filed for IPO in recent weeks — OpenAI at a reported $1T+ valuation, Anthropic at $965B. The CEOs will join G7 heads of state for a working lunch, in what analysts are calling a new era of “algorithmic diplomacy.”

The backdrop: the Hiroshima AI Process launched at the 2023 G7, and since then, AI governance has moved from a side topic to a central pillar of international diplomacy.

🔗 Mercury News — Full story | TNW — Background

The AI Price War Is Here: Enterprises Ditch Premium Models for Cost-Efficient Stacks

The WSJ and IBTimes both report this week that businesses are abandoning premium AI models from OpenAI and Anthropic in favor of cheaper alternatives — not out of ideology, but cold economics. Companies are building hybrid stacks: routine tasks (summarization, classification) go to low-cost or open-source models, while premium systems are reserved for complex reasoning.

Accenture’s 2026 analysis shows dynamic routing between models is delivering 40-60% productivity gains. The market is fragmenting fast — open-source models like Llama and regional players from China and Europe are competing on price, and cloud providers (AWS, Azure, Google Cloud) are the quiet winners as enterprises diversify.

For SMBs, this is good news. The commoditization of AI means you no longer need a $200/month per-seat license for every use case. Smart routing between models — premium for strategy, budget for triage — is the playbook that’s emerging.

🔗 IBTimes — Full story | WSJ — Analysis

🚀 Want AI working for YOUR business? Most companies are experimenting with AI chatbots. We deploy AI workforces — AI Employees that follow up on leads, resolve support tickets, publish content, chase invoices, and screen 200 job applicants overnight so your hiring manager starts Monday with the top 10. Each role has a cost profile and human oversight, managed through one platform. This newsletter? Written by an AI Employee, approved by a human — so our team stays focused on what only humans can do. AIToken Labs helps businesses design their AI Workforce Operating Model — starting with the 2-3 roles that deliver ROI in the first 60 days. Book a free 40-minute AI Workforce Blueprint Session →