Good morning — it’s Wednesday, July 8. Today’s lead story isn’t about a shiny new model. It’s about the quiet crisis unfolding inside companies that already bought into AI agents — and why most of them are heading for a wall. Let’s dig in.

🤖 AI AGENTS

Gartner Drops a Bombshell: Over 40% of Agentic AI Projects May Be Canceled by 2027

July 7, 2026 — Forbes / Gartner

Here’s a stat that should make every AI buyer pause: more than 4 out of 10 agentic AI projects currently underway may be scrapped within the next 18 months. Not because the models aren’t good enough — but because the organizations deploying them aren’t ready.

Gartner’s latest forecast paints a sobering picture. The culprit isn’t model capability. It’s governance, data access, and deployment discipline. Agents that perform beautifully in controlled demos hit a wall when they encounter messy real-world systems — missing fields, duplicated records, permission walls, outdated workflows.

Meanwhile, the UK AI Safety Institute analyzed 177,000+ agent tools and found “action” tools — agents that send emails, change files, and move money — jumped from 24% to 65% of all agent usage in just 16 months. The stakes are rising fast.

So what? The companies that survive the agentic AI wave won’t be the ones with the biggest models. They’ll be the ones that can answer three questions before deploying: (1) What’s the success metric and who agreed to it? (2) Does the agent actually have access to the data and tools it needs — today? (3) Who notices when it fails, and how fast can you roll back?

Read the full analysis on Forbes →

⚡ Section Report: Less Than 10% of Employees Are Ready to Use the AI Agents Their Company Deployed

July 7, 2026 — Section AI Proficiency Report

69% of workers say their organization has deployed some form of AI agent. But here’s the problem: 82% can’t correctly identify what an AI agent is on a multiple-choice test. Only 5.5% of the workforce meets the AI proficiency bar. Even more alarming: only one-third of employees at “agentic” organizations have received any training. C-suite executives are 5x more likely to have received agent training than individual contributors. The deployment gap isn’t just technical — it’s a training crisis.

Read the full report →

⚡ Peraton Launches Agentic AI Platform for Government’s Most Sensitive Missions

July 8, 2026 — Peraton

Government contractor Peraton unveiled Peraton[x], a “full-spectrum composable AI platform” purpose-built for mission-critical government operations. The platform ingests and reasons across mixed enterprise data types in real time — think intelligence analysis, defense logistics, and emergency response coordination. CEO Steve Schorer calls it a platform that “doesn’t replace humans” but fundamentally changes how agencies scale. With defense and intelligence budgets pouring into AI, this signals where the federal agent market is heading: secure, on-prem, and mission-grade.

Learn more about Peraton[x] →

📰 AI NEWS

OpenAI GPT-5.6 Goes Public Thursday — Three Models, Half the Cost of Claude

July 8, 2026 — Reuters / OpenAI

After weeks of government-gated previews, OpenAI confirmed GPT-5.6 will launch publicly on Thursday, July 9. The release includes three tiers: Sol (flagship), Terra (mid-range for everyday work), and Luna (fast, low-cost). Early reports peg Sol’s pricing at roughly half the cost of Anthropic’s Claude Fable 5 — a direct shot across the bow in the inference pricing war. The Trump administration greenlit the broad launch after security testing. Worth noting: both GPT-5.6 and Anthropic’s Mythos series have drawn concern over “unprecedented ability to identify software vulnerabilities.”

Read more →

Microsoft Quietly Swaps OpenAI and Anthropic for Its Own Models in Excel and Outlook

July 7, 2026 — Bloomberg / TechCrunch

Microsoft is routing thousands of AI prompts in Excel, Word, and Outlook to its in-house MAI (Microsoft AI) models instead of OpenAI and Anthropic — a move driven entirely by cost. After launching seven MAI models at Build in June, Microsoft is now executing the switch in production. Anthropic’s models remain for select Office use cases, but the direction is clear: the hyperscalers are building their own engines. Amazon, Uber, Meta, and Accenture are all pursuing similar cost-cutting strategies. The message for businesses? Don’t get too comfortable with any single model provider — the landscape is fragmenting fast.

Read the TechCrunch report →

SambaNova Hits $11B Valuation — AI Chip Startup Was Almost Bought for $1.6B

July 8, 2026 — CNBC / Bloomberg

SambaNova just closed the first tranche of a $1 billion Series F at an $11 billion post-money valuation — a staggering number for a company Intel reportedly tried to acquire for just $1.6 billion months ago. JPMorganChase is already using SambaNova’s RDU chips for on-prem AI inference. CEO Rodrigo Liang says the company is “strongly considering” an IPO. The takeaway: the AI chip race isn’t just NVIDIA vs. everyone else anymore. Inference-specific silicon is becoming its own category — and the money is pouring in.

Read more →

🚀 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. →