ADVERTISE | PODCASTS | EXECUTIVE’S PASS | B2B TRAINING | | | Hey | Welcome to The AI Report — the #1 AI newsletter for business leaders. | You don’t need to be technical. Just informed. | • 1. 🦾 Build high-end Claude skills with Innovating with AI
• 2. ⚡️ AI prices hit new record lows
• 3. 💼 Your Business Briefing
• 4. 🔐 Reclaim your privacy today with Incogni
• 5. 🤝 The Human & AI De Brief with Marie-Lou
• 6. ✍️ Today’s Policy Corner
• 7. 🗞️ The News Bulletin | Are you ready to launch AI transformation for your company? | |
|
| | TOGETHER WITH INNOVATING WITH AI | | Want to get more out of Claude? Build personalized skills (step-by-step guide) | | You're paying for Claude but not hitting your full potential. This free guide shows you how to build high-end Claude Skills with no prior tech knowledge required. Your style, your voice, your data, supercharged. | |
|
| | Latest in AI | | | AI prices hit new record lows | 🚨 Our Report | The cost of running AI models has fallen to record lows, with average inference prices hitting $1.16 to $1.18 per million tokens in early August 2026. That's a 43% drop from late May and roughly 12% of March 2023 pricing levels. | 🔓 Key Points | OpenAI slashed GPT-5.6 Luna pricing by 80% in late July, dropping input tokens to $0.20 per million and output tokens to $1.20 per million. Chinese competitors like DeepSeek have achieved cuts up to 99%. ARK Invest calls it a "virtuous cycle": lower prices drive adoption, higher volumes justify infrastructure spend, and that enables further cuts. Token volumes have surged tenfold in some agent-driven applications. Comparable AI capabilities now cost over 280 times less than early 2023, with hardware efficiency gains compounding the commercial price wars. Enterprise automation use cases that were marginal at $2 are now obvious investments.
| 🔐 Relevance | For enterprise buyers, this repricing fundamentally changes the math on AI deployment. Agent-heavy workflows that chain hundreds of model calls per interaction become viable at current rates. The combination of OpenAI's aggressive cuts and Chinese open-weight competition signals that inference costs will keep falling, making "wait and see" increasingly expensive as competitors automate first. | |
|
| | | | THE BUSINESS BRIEFING: IT & OPERATIONS (powered by Upscaile) | | An 85,000-employee global retail bank was drowning in 26,000 IT tickets a month. 68% were the same L1 questions on repeat: password resets, MFA enrolment, VPN access. Mean-time-to-first-response sat at 47 minutes during business hours and over 4 hours overnight. They deployed a hierarchical multi-agent platform on LangGraph + GPT-4o embedded directly inside ServiceNow ITSM. | Tool used: LangGraph + OpenAI GPT-4o + Microsoft AutoGen -- multi-agent IT service desk automation with native ServiceNow integration. | Result: 18,000 L1 tickets auto-resolved monthly. Mean-time-to-first-response collapsed from 47 minutes to under 60 seconds. $1.8M annualised savings in year one. Zero SOX audit findings. | The lesson: The bank didn't bolt AI onto a broken workflow. They embedded agents inside the existing ServiceNow console so L1 techs never had to switch tools. SOX-controlled and BSA/AML-flagged tickets always route to humans with full agent transcripts attached. Compliance stayed intact because the architecture assumed escalation, not autonomy. | Steal this: Map your top five L1 ticket categories this week. If 60%+ are password, VPN, or access requests, you're sitting on automation gold. Start with one category, one integration (Okta or AD), and a hard rule that anything privileged routes to a human. Build confidence before you build scale. | |
|
| | TOGETHER WITH INCOGNI | | Remove your personal data from Google search results | Your phone number, address, DOB, even your SSN can show up in a Google search. Marketers buy it, scammers exploit it, and in the worst cases, it ends up for sale on the dark web. | Incogni Unlimited helps you take control. Our privacy agents hunt down and remove your personal data from an unlimited number of sites. With automation across 420+ broker types—including People Search platforms and private databases—Incogni erases your information where others can’t reach. | Protect your identity before it’s too late. Use code AIREPORT for an exclusive 55% discount on unlimited removals. | |
|
| | THE HUMAN & AI DE BRIEF | | Three of your collaborators’ biggest AI fears are already management problems | | The four most commonly expressed concerns: unreliability, job anxiety, loss of autonomy and agency, and cognitive atrophy are not isolated worries, they compound. The person worried about unreliability is often the same person worried about their job, their thinking, their conclusions. | Three of those four fears are already management problems. The fourth requires something transparent leadership hasn’t been asked to do yet. Read the full breakdown here. | |
|
| | THE POLICY CORNER | | Japan mandates human checkpoints for AI agents that take external actions. | The AI Operator Guideline v1.2, jointly issued by METI and MIC on March 31, 2026, elevates Human-in-the-Loop from "encouraged" to an implementation-level requirement for any AI system that sends emails, writes to external APIs, executes payments, pushes code to production, or controls physical devices. The rule applies to developers, providers, and business users of AI operating in Japan. | Deadline: In effect now. No grace period for new deployments. | Risk: Soft law with hard consequences. Ministries reference the guideline as the baseline for procurement criteria and security assessments. Non-compliance surfaces during contract negotiations, vendor audits, and IPO prep. | Your move: Build an inventory of every AI system touching external resources this week. Map each to its approval layer: prompt-level, plan-level, or action-level. High-risk actions (money, production environments, outbound comms) need all three checkpoints. Document who approves, when, and retain logs that prove human judgment was in the loop. | |
|
| | AI News | | 🔓 OpenAI's Astra model hits "critical cybersecurity threshold": The forthcoming model scored perfect on ExploitBench and discovered two zero-day vulnerabilities without human guidance, prompting OpenAI to restrict access to its most advanced hacking capabilities. FULL STORY 💰 Anthropic cuts costs with Fable 5.1 release: The new model ships with zero data retention for enterprise clients and fewer false-positive restrictions, while Mythos 5.1 remains locked to registered cybersecurity and life sciences partners. FULL STORY 🍎 Tim Cook steps aside as Apple CEO: John Ternus takes over a $4.5 trillion company with AI as his top challenge after weak Vision Pro sales and Siri delays left Apple playing catch-up to rivals. FULL STORY 🎨 Google launches Pics to challenge Canva and Adobe Express: The AI-first design tool, powered by Google's Nano Banana model, rolls out to Workspace customers and Google AI subscribers starting in Docs and Slides. FULL STORY
| |
|
| | | Trending AI Tools (Sponsored by the AI Executive’s Pass) | | A curated look at the AI tools quietly transforming how teams work. | Gamma* — Create stunning presentations and websites from a single prompt, no design skills required AISQ is an AI marketing platform for automating SEO, content, social media, email, and digital growth workflows. Wegic is an AI website builder for creating and editing websites through simple natural-language conversations.
| | *indicates a sponsored tool, if any | | ⚡️ Why pay for 4–6 separate AI tools at full price when the AI Executive’s Pass fixes that? | |
|
| | The Money: Agent infrastructure security emerges as enterprise priority | | Two of this week's largest AI seed rounds targeted the same thesis: as autonomous agents spread across enterprise systems, the security and reliability layers around them become mission-critical. Sequoia backed both bets within weeks, signaling conviction that agent oversight is the next infrastructure category. | Deals to know: | Empirik (Seed, $21M) -- Predicts infrastructure outages before they occur by tracking system changes and inferring ripple effects. Already deployed at S&P Global and Guardant Health. Investors: Sequoia, Canapi, Alumni Ventures AIR (Seed, $50M across two rounds) -- Discovers agents running inside companies and continuously vets the skills, plug-ins, and MCP servers they use. 20+ customers, quarter are large enterprises. Investors: Greenoaks (lead), Sequoia, Wiz co-founder Yinon Costica, Clay co-founder Varun Anand
| Signal: Capital is flowing to companies that make agents enterprise-ready, not to agents themselves. The pattern suggests VCs expect agent adoption to accelerate faster than security tooling can mature organically. |
|
| | Thoughts on today's edition? | |
| |
|
|
| | Refer a Friend | | Latest episode: What Happens When You Give an AI Agent a Shopping Budget? → Listen here
Want to reach 400k+ decision-makers? → Sponsor us | Until next time, Arturo and Liam. | | P.S. Unsubscribe if you don’t want us in your inbox anymore. |
|
| | | | | | | | What did you think of this edition? | |
| |
| |
| | |
|
|
