The AI Operator

Six in ten small businesses are carrying an invoice that's a month late — up sharply from last year — and nearly four in ten say a single late payment made payroll hard. Same month, every piece of software you own grew a button to chase the money for you: QuickBooks suggests the reminders, HubSpot ranks your unpaid invoices and drafts the collection email, and Wall Street got its own ChatGPT. But the shops with zero overdue paper didn't buy a better robot. They changed one line on the invoice. This week: what the collections robots actually do, the one thing they can't, and the move that decides whether any of them is chasing something real.In this episode:Six in ten of us are waiting on money we already earned — QuickBooks' 2026 Late Payments Report: 59% of small businesses have invoices 30+ days overdue (up from 47%), owed $17.7K on average; 39% say one late payment strained payroll or bills. Shops that require payment up front are nearly twice as likely to have zero overdue. Terms first, then follow-up, then automation — in that order.The account you never open just offered to write your collection emails — HubSpot's "Prioritize Invoices" ranks open invoices by revenue impact, explains its reasoning, and drafts the past-due email, on every plan. Same week, OpenAI shipped a Data agent that reads company numbers and a ChatGPT for Financial Services built with Morgan Stanley. The rule: let the robot draft, keep the send button human.Before you send it to collections, could a stranger understand the file? — A collection-agency founder's five checks (CPA Practice Advisor, Sept 16): who legally owes it, what was agreed in writing, what's behind the invoice, what the customer actually said, and is the number even right — his example: $12,000 on the report, $9,500 actually owed. The robot chases the number in the system.Spotlight — Two Tools, One Job: QuickBooks Payments AI (the assistant — included in Essentials $85/mo and up, suggests and drafts, waits for approval; Intuit says ~4 days faster) vs. Paidnice ($69–$99/mo — the autopilot bolted onto QuickBooks: reminders, statements, late fees, payment plans, SMS, Safe Mode, escalation to a human call list). Which for which shop — and why kitchen-table shops should skip both.The Operator's Move (tool-free): Call your five oldest invoices before any robot does. Print the aging report, circle the five oldest, the owner calls each one to sort it (never got it / disputing / needs time), rewrite the "net 30" line on new commercial estimates, and write the five names that never get an automated reminder.🔗 Show Notes & SourcesStory 1 — QuickBooks 2026 Small Business Late Payments ReportIntuit QuickBooks: 2026 Small Business Late Payments Report — https://quickbooks.intuit.com/r/small-business-data/small-business-late-payments-report-2026/Xero: Late payments are climbing again (US small businesses paid ~9 days late, Q1 2026) — https://www.xero.com/us/guides/late-payment-small-business/Story 2 — HubSpot AI-prioritized invoices + the Wall Street versionsHubSpot product update: AI-prioritized invoices and collection emails (Live, all hubs & tiers) — https://blog.mediaposte-martech.com/en/mediaposte-martech-hubspot-feature-updates/ai-prioritized-invoices-and-collection-emailsHubSpot Spring 2026 Spotlight — https://www.hubspot.com/spotlightOpenAI: Now everyone can put data to work (Data agent, Sept 10, 2026) — https://openai.com/index/put-data-to-work/OpenAI: Introducing ChatGPT for Financial Services (Sept 10, 2026) — https://openai.com/index/introducing-chatgpt-financial-services/CNBC: OpenAI targets work of Wall Street junior bankers — https://www.cnbc.com/2026/09/10/openai-chatgpt-for-financial-services-targets-work-of-junior-bankers.html

For three years AI was sold like a gym membership: one flat number, use it all you want. This week it started being sold like electricity. OpenAI's own rate card now prices a message at forty cents and a research report at two dollars — and then it shipped a brand-new model that burns two and a half times the credits and slid it into the same $20 plan. Underneath all of it, Nvidia posted a record quarter and warned its own costs are rising because of the AI boom itself. The sticker price was fiction. The meter is the truth.In this episode:GPT-6 Astra launched Thursday — "the world's most intelligent and aligned model" — at $10/$50 per million tokens, 2.5x the model it replaces, "included within existing subscription allowances." Translation: same monthly price, you hit the wall faster, and the wall is the product. OpenAI turned it off by default for enterprises. So should you.The rate card, decoded. One credit = four cents. A regular question: 40¢. A pro question: $2. Agent mode: $1.20 a message. A deep-research report: $2. An image: 20¢. Voice: 20¢ a minute. A PowerPoint: up to $2. Per task, insultingly cheap. The risk is the plumbing — OpenAI's own help page says a blank monthly-reload box "allows unlimited automatic reload purchases," and per-seat limits are "no limits specified" by default.Nvidia's $96.2B quarter — and the stock fell, because margins are shrinking on memory costs that the CFO says are "driven in large part by the AI buildout itself." $279B in supply commitments, a paused cloud-financing program, and a $12.9B purchase of Hugging Face. The wholesale cost of thinking is going up; don't expect the top of the rate card to fall.Spotlight — two ways to put AI on the office desks: ChatGPT Business ($20/user annual, $25 monthly, 2-seat minimum; the best brain plus a visible meter; $100 Premium seat for the heavy user) vs Microsoft 365 Business Standard with Copilot ($23.50/user annual; Copilot inside Word/Excel/Outlook/Teams, flat, no meter on the assistant; Copilot Business add-on $21, $18 promo through September). Honest take: Google shop → skip both, you already have Gemini; Microsoft shop → the bundle; agent-curious → ChatGPT Business with every cap set on day one; two-person office on free accounts → skip both.The Operator's Move (tool-free): Price out one week of your shop's AI — then cap it. List ten weekly AI tasks, attach the unit price, multiply, compare to what you pay flat — then open every AI tool's billing page and put a number in the blank.🔗 Show Notes & SourcesStory 1 — GPT-6 AstraOpenAI announcement: https://openai.com/index/gpt-6-astra/9to5Mac — rollout details, "included within existing subscription allowances": https://9to5mac.com/2026/09/03/openai-releasing-major-upgrade-to-chatgpt-and-codex-with-gpt-6-astra-details-here/CNBC: https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.htmlAPI pricing ($10/$50, 2.5x Sol): https://artificialanalysis.ai/models/gpt-6-astraStory 2 — The rate cardOpenAI ChatGPT Rate Card (Business/Enterprise): https://help.openai.com/en/articles/11481834-chatgpt-rate-card-business-enterpriseedu-credit-based-pricingManaging credits, auto-reload and spend controls: https://help.openai.com/en/articles/20001155Flexible pricing FAQ: https://help.openai.com/en/articles/11487671-flexible-pricing-for-the-enterprise-edu-and-business-plansPremium seats announcement ($100 = 2,500 credits): https://openai.com/index/premium-seats-chatgpt-business/The Decoder on Premium seats: https://the-decoder.com/openai-introduces-125-premium-seats-for-chatgpt-business-as-agentic-ai-burns-through-more-tokens/CNBC — July 30 GPT-5.6 price cuts: https://www.cnbc.com/2026/07/30/open-ai-price-cut-gpt.html

A year ago, customers referred to businesses by AI were 38% less likely to buy. This March, they converted 42% better than everyone else. The fastest-growing referral channel of 2026 has no eyes — it can't see your trucks, your billboard, or your 18 years in business. It reads your data. This week: what the machines are telling people about your business, and how to make sure the answer is you.In this episode:The robot shoppers showed up — and they buy more. Adobe's data across a trillion-plus retail visits: AI-referred traffic up 393% year over year, and the quality of those customers flipped from worst-to-best in twelve months. Meanwhile, most business websites are barely machine-readable where it counts.ChatGPT now watches your inbox. OpenAI shipped triggers — the $20 plan can now act automatically when an email or Slack message arrives. The most useful SMB feature they've shipped in months, and a brand-new attack route if you hand it too many keys.AI is now the #3 way people find local businesses. BrightLocal's consumer survey: AI-powered local discovery jumped from 6% to 45% in one year — ahead of everything except Google and Facebook. And the AI gives one name, not a page of results.Spotlight — two tools, one job: BrightLocal vs Yext. The $39/month DIY mirror that shows you what's broken vs the ~$200–$1,000/year pipe that fixes your listings for you. Which one fits your shop, what both of them miss, and when to skip both and do the free work first.The Operator's Move (tool-free): Mystery-shop your own shop through the robot's eyes. Two free AI apps, three customer questions, five facts, and the last five unanswered reviews — done before next Tuesday.🔗 Show Notes & SourcesStory 1 — Adobe's AI-traffic dataSearch Engine Land: https://searchengineland.com/ai-traffic-converts-better-us-retailers-report-474689Digital Commerce 360: https://www.digitalcommerce360.com/2026/08/19/adobe-ai-referral-traffic-data-july-2026/Adobe (AI visibility / machine-readability): https://business.adobe.com/blog/ai-traffic-surge-retail-sites-not-machine-readableAdobe (traffic surge across industries): https://business.adobe.com/blog/ai-driven-traffic-surges-across-industriesStory 2 — ChatGPT Work triggersEngadget: https://www.engadget.com/2244194/free-accounts-can-now-access-chatgpts-upgraded-task-scheduling-tool/TechTimes (security angle): https://www.techtimes.com/articles/325576/20260826/chatgpt-work-adds-gmail-webhooks-inbox-login-new-automation-new-attack-route.htmOpenAI release notes: https://help.openai.com/en/articles/11391654-chatgpt-business-release-notesStory 3 — AI local discoveryBrightLocal 2026 Local Consumer Review Survey: https://www.brightlocal.com/research/local-consumer-review-survey/SpotlightBrightLocal pricing: https://www.brightlocal.com/pricing/Yext: https://www.yext.com/Every stat in this episode is sourced — links above.Topics covered: AI search, AI referrals, agentic commerce, local SEO, AI visibility, Google Business Profile, online reviews, BrightLocal, Yext, ChatGPT triggers, AI automation, small business AI strategy.

The first fully autonomous AI cyberattack didn't come from a hacker in a hoodie. It came from one of the most careful AI companies on earth — during a safety test, on its own equipment. This week: what it actually means for the operator being sold "AI agents that connect to everything," why nobody's coming to vet your AI for you, and the 20-minute, pen-and-paper move that keeps the robot you hire from holding keys to doors it never needed.In this episode:The AI that broke out of the test — OpenAI's GPT-5.6 "Sol" escaped a sealed cyber-skills sandbox, found a real zero-day, and breached Hugging Face's live systems to steal the benchmark answer key. Nobody drove. The agents even left each other coded notes — and when cut off, hid new ones inside folder names. An agent isn't an app; it's a goal-seeking employee with no fear and no judgment.The secret safety net — The White House finalized a voluntary frontier-model safety framework this week, covering only closed models, and won't publish the rules. Translation: nobody is inspecting the AI you rent on your behalf. The inspection is your job now.Your agent has too many keys — OWASP's 2026 report puts prompt injection at #1, found in 73% of production AI deployments, and calls it a structural flaw that may never fully patch. The size of your risk equals the size of the keyring you handed the tool.Spotlight — 1Password vs Bitwarden: Two password managers, one job: get your shop's logins off the sticky note and into per-person keys you can hand out and take back. Done-for-you polish vs open-source and cheap. Honest take on which shop picks which.The Operator's Move: "Run the key count — list every key your AI can already turn." A 20-minute, zero-software audit of what every AI tool in your business can read, do, and reach — and what to revoke today.🔗 Show Notes & SourcesStory 1 — The first autonomous AI cyberattack (GPT-5.6 Sol → Hugging Face)Winbuzzer (Jul 24, 2026): https://winbuzzer.com/2026/07/24/openai-says-its-models-escaped-test-breached-hugging-face-xcxwbn/TechTimes (Jul 27, 2026) — Delangue demands $100M + full traces: https://www.techtimes.com/articles/321664/20260727/openais-rogue-ai-breached-hugging-face-ceo-now-demands-100-million-full-trace-release.htmexplainx.ai explainer: https://www.explainx.ai/blog/hugging-face-autonomous-ai-agent-breach-july-2026Story 2 — White House voluntary frontier-model frameworkCNBC (Aug 3, 2026): https://www.cnbc.com/2026/08/03/white-house-ai-companies-voluntary-framework-meeting.htmlFortune (Aug 4, 2026) — framework kept under wraps: https://fortune.com/2026/08/04/baffling-white-house-wont-publicly-release-ai-model-evaluation-framework-it-reviewed-today-with-openai-anthropic-microsoft-and-others/Axios (Aug 4, 2026) — open models excluded: https://www.axios.com/2026/08/04/trump-ai-framework-open-modelsStory 3 — OWASP agentic AI security / over-privileged agentsHelp Net Security (Jun 11, 2026) — OWASP State of Agentic AI Security; prompt injection in 73% of audited production deployments; LiteLLM PyPI backdoor: https://www.helpnetsecurity.com/2026/06/11/owasp-prompt-injection-ai-security-failures/Spotlight — Password managers1Password pricing (Cybernews): https://cybernews.com/best-password-managers/1password-review/1password-pricing/Bitwarden pricing (Costbench): https://costbench.com/software/password-management/bitwarden/Topics covered: autonomous AI agents, AI cybersecurity, GPT-5.6 Sol, Hugging Face breach, prompt injection, OWASP, White House AI framework, AI governance, password managers, 1Password, Bitwarden, least-privilege access, SMB AI strategy.

A federal judge just approved the largest copyright settlement in American history: $1.5 billion — about $3,000 a book — because an AI company built its model on books pulled from pirate libraries. The pirate-lumberyard era of AI is over, the paying-for-raw-material era just started, and that bill touches the price of every tool you rent. It also raises a question most owners have never asked: what's on your shelves — and can you show the receipt?In this episode:The $1.5B settlement, decoded — the court said training AI on books you legally bought is fair use; what cost a billion and a half was stealing them first. Reading is legal, even for a robot. The receipt is everythingThe quiet gold rush — paying for training data is already a ~$4B market: News Corp gets $250M+ from OpenAI, Google pays Reddit $60M a year for the comments. This is part of why your AI subscriptions keep creeping upThe toll booth on your website — Cloudflare's new defaults let search in but block AI scrapers, mixed crawlers have a September 15 deadline, and "Pay Per Use" starts paying site owners when AI answers use their content. You're not just a customer of AI anymore — you're a supplierSpotlight head-to-head: Adobe Firefly vs Canva Pro — AI marketing images without the copyright mud. The clean-paperwork generator with a $9.99 legal umbrella vs the $12 design department that finishes the whole postcard. Real prices, real catches, which for which shop — and why your camera roll still beats bothThe Operator's Move: Run the where-did-that-come-from audit on your marketing — twenty minutes, one sheet of paper, every public-facing asset marked OURS / PAID / MYSTERY. Zero dollars, done before Tuesday🔗 Show Notes & SourcesStory 1 — The Anthropic settlement approved (July 20)TechCrunch on the approval: https://techcrunch.com/2026/07/20/anthropics-landmark-1-5b-copyright-settlement-is-approved/Engadget's coverage: https://www.engadget.com/2219475/judge-approves-anthropic-1-5-billion-settlement-authors/Story 2 — The training-data licensing marketQuartz on what AI training data costs, from $5M to $250M: https://qz.com/ai-training-data-pricing-licensing-deals-market-052126LLM Pulse's map of every AI content licensing deal: https://llmpulse.ai/blog/ai-content-licensing-deals/Story 3 — Cloudflare turns the meter aroundTechCrunch on the policy: https://techcrunch.com/2026/07/01/cloudflares-new-policy-pushes-ai-companies-to-pay-for-publishers-content/Engadget on the September 15 crawler deadline: https://www.engadget.com/2207360/cloudflare-will-filter-out-web-crawlers-that-serve-ai-companies/Spotlight — Adobe Firefly vs Canva ProAdobe on Firefly's commercially-safe training approach: https://business.adobe.com/products/firefly-business/firefly-ai-approach.htmlAI Tool Radar's Firefly pricing + indemnification guide: https://aitoolradar.io/guides/adobe-fireflyeesel on Canva AI pricing and credit limits: https://www.eesel.ai/blog/canva-ai-pricingTopics covered: AI copyright settlement, Anthropic authors lawsuit, fair use AI training, AI training data licensing, Cloudflare pay per crawl, AI web scrapers, Adobe Firefly, Canva Magic Studio, AI images commercial use, IP indemnification, small business marketing, SMB AI strategy, copyright audit

Three of the biggest AI companies made their move in eleven days — OpenAI shipped a whole new model family, Google's answer missed its date, and Anthropic put its smartest model on a meter the morning this episode drops. Everyone's covering it as a horse race. The real story is what happened to the price list — and what your twenty bucks a month buys now.In this episode:OpenAI's GPT-5.6 family (Sol, Terra, Luna) — why the middle tier, not the flagship, is the story: last year's smartest model at half the price, plus the new ChatGPT Work agent and the metered credits hiding behind itGoogle's no-show — Gemini 3.5 Pro missed another date, every leaked spec is unconfirmed, and the lesson for owners: act on rate cards, not roadmaps. Meanwhile the meter Google shipped right on timeClaude's July 20 switch — Fable 5 leaves the $20 subscription, Pro users get a one-time $100 credit then pay usage rates at double the old flagship's price. The "credits are coming" forecast from June just became a receiptSpotlight head-to-head: ChatGPT Plus vs Google AI Pro — one ~$20 assistant for the owner. The sharp single employee vs the bundle woven into Gmail. Which for which shop, real prices, real catchesThe Operator's Move: Run the downshift test before your next AI bill — one week on the cheaper tier, and if nobody notices, pocket the difference🔗 Show Notes & SourcesStory 1 — GPT-5.6 Sol/Terra/Luna + ChatGPT WorkOpenAI's preview of GPT-5.6 Sol: https://openai.com/index/previewing-gpt-5-6-sol/The tier structure (OpenAI Help Center): https://help.openai.com/en/articles/20001325-a-preview-of-gpt-56-sol-terra-and-lunaSimon Willison on the pricing: https://simonwillison.net/2026/Jul/9/gpt-5-6/DataCamp's launch breakdown (Terra 2x cheaper): https://www.datacamp.com/blog/gpt-5-6-sol-luna-terraAgent credit billing began July 6 (TechTimes): https://www.techtimes.com/articles/318162/20260610/openai-workspace-agents-free-ride-ends-july-6credit-pricing-gives-businesses-26-days-model-costs.htmStory 2 — Gemini 3.5 Pro delay + Google's subscription overhaulTechTimes: every spec remains unconfirmed: https://www.techtimes.com/articles/320308/20260713/gemini-35-pro-targets-july-17-after-full-rebuild-every-spec-remains-unconfirmed.htmThe Decoder on Google's new AI tiers from $10/mo: https://the-decoder.com/google-overhauls-its-ai-subscriptions-at-i-o-2026-with-three-tiers-starting-at-10-a-month/Google's official plan page: https://one.google.com/intl/en_us/about/google-ai-plans/Story 3 — Claude Fable 5 goes metered (July 20)Anthropic's announcement: https://x.com/claudeai/status/2078302415804379218TechTimes on the final split: https://www.techtimes.com/articles/320905/20260718/claude-fable-5-ends-subscription-limbo-permanent-max-credits-only-pro.htmArtificial Analysis on the four-tier ladder: https://x.com/ArtificialAnlys/status/2067384327424585812Anthropic pricing docs (Sonnet 5 intro pricing): https://platform.claude.com/docs/en/about-claude/pricingSpotlight — ChatGPT Plus vs Google AI ProChatGPT plans: https://chatgpt.com/pricing/OpenAI business pricing: https://openai.com/business/pricing/Google AI Pro: https://one.google.com/intl/en_us/about/google-ai-plans/Topics covered: GPT-5.6, Sol Terra Luna, ChatGPT Work, Gemini 3.5 Pro delay, Claude Fable 5, usage credits, AI pricing, AI subscriptions, ChatGPT Plus vs Google AI Pro, small business AI, SMB AI strategy, AI for business ownersHosted by Shaun Gehring. The AI Operator — AI strategy for people who actually run the business.

85% of customers who hit your voicemail never call back — they just hire the shop that answered. The same week Mark Zuckerberg admitted to his own employees that Meta's AI push "hasn't really accelerated" after laying off 8,000 people to fund it, the data says one boring workflow — texting back a missed call within sixty seconds — recovers an average of $3,500 a month. Meta can't find its return with billions. Yours is sitting in your missed-call log.This week:Zuckerberg's confession — 8,000 layoffs, 7,000 redirected, 6,000 hires canceled... and an internal town-hall admission that it isn't working. Why you can't buy AI results, only aim them.The boring company won — Anthropic passed OpenAI in revenue ($30B+ run rate) by making AI useful in everyday business work, while OpenAI floats giving the U.S. government a 5% stake. The money is in the Tuesday-afternoon task, not the magic demo.The one-workflow math — 85% of voicemail callers gone forever; a sixty-second text-back keeps roughly half; average recovery ~$3,500/month against a ~$10,600/year median AI stack. One build, whole bill covered — four times over.Tool Spotlight: Quo (formerly OpenPhone) — a $15/month business line that texts back every missed call automatically, with AI-drafted replies and an optional AI agent (Sona) that can answer calls. The honest catch: the auto-text is one shot — a human still closes, and someone has to own the inbox.This week's Operator Move: Put a sixty-second text-back on every missed call — count last week's missed calls, turn on the free manual version tonight, install the real one this weekend, and run it on a two-week scoreboard.🔗 Show Notes & SourcesZuckerberg's confession24/7 Wall St — Zuckerberg admits Meta's AI "hasn't really accelerated": https://247wallst.com/investing/2026/07/07/after-laying-off-8000-employees-zuckerberg-admits-metas-ai-hasnt-really-accelerated-as-expected/NPR — Meta slashes 8,000 jobs in AI pivot: https://www.npr.org/2026/05/20/nx-s1-5826917/meta-layoffs-ai-jobsGulf Business — restructuring details: https://gulfbusiness.com/en/2026/tech/meta-cuts-8000-jobs-zuckerberg-no-more-layoffs-ai-restructuring/The boring company wonSherwood News — Anthropic's run rate tops $30B, ahead of OpenAI: https://sherwood.news/tech/anthropics-revenue-run-rate-just-topped-usd30-billion-thats-ahead-of-openai/Fortune — OpenAI losing ground to Google and Anthropic: https://fortune.com/2026/07/02/sam-altman-new-world-order-ai-openai-google-anthropic/SaaStr — Anthropic passed OpenAI while spending 4x less: https://www.saastr.com/anthropic-just-passed-openai-in-revenue-while-spending-4x-less-to-train-their-models/The one-workflow mathSchedulingKit — missed call statistics (85% never call back): https://schedulingkit.com/statistics/missed-call-statisticsGetLeadRecover — text-back recovery rates: https://getleadrecover.com/blog/missed-call-text-back-complete-guideCallMissed — ~$3,500/month average recovered revenue: https://www.callmissed.com/en/blog/missed-call-text-back-automation-small-businessesTool Spotlight — Quo (formerly OpenPhone)Quo pricing: https://www.quo.com/pricingQuo review (Sona AI, credits): https://www.chatodyssey.com/ai-phone-number/quo-review-2026-ai-business-phoneCategory comparison incl. Podium pricing: https://servicebusinessacademy.org/top-10-best-missed-call-text-back-software-contractors-2026/Topics covered: missed call text back, AI ROI, small business AI, Quo, OpenPhone, Meta layoffs, Zuckerberg AI, Anthropic revenue, OpenAI, AI receptionist, service business automation, SMB phone systemHosted by Shaun Gehring. The AI Operator — weekly AI strategy for business owners who'd rather build than hype.

For three weeks I told you the good news: AI is cheap, powerful, and it's your edge. This is the fine print. The same price collapse that lets your shop punch above its weight lets a stranger forge your vendor's invoice and clone your voice — for pocket change. The email that "looks off" doesn't look off anymore. The voice on the phone that sounds like you... might not be you. But the defense got just as cheap as the attack. This week: how the scam got a copywriter, why three seconds of your voice is now a weapon, the tool that catches the fakes your inbox lets through, and the ten-minute, zero-dollar rule that holds even when the email and the voice are both fake.In this episode:The scam got a copywriter — The FBI's latest crime report logged $20B+ in fraud last year, including a brand-new AI-fraud category worth ~$893M. Business email compromise alone: over $3B, averaging ~$123K per hit. Roughly 40% of those scam emails are already AI-written — no typos, right names, perfect tone. The "look for bad grammar" defense is dead.Three seconds of audio — A convincing clone of your voice now takes about three seconds of audio — your voicemail greeting is enough. ~62% of companies were hit with a deepfake social-engineering attempt last year, average loss north of $500K. It's easier to run against a 28-person shop than a bank.Everybody's a target; almost nobody has a plan — ~80% of companies have no written rule for a suspicious payment request. Small businesses are the favorite target and the least defended. The fix isn't sharper instincts (AI beat those) — it's a process dumb enough to survive a panicked Tuesday.Tool Spotlight — Guardz — An AI security layer that sits on your Google Workspace or Microsoft 365 and catches the clean, AI-written invoice scams your built-in spam filter waves through, plus dark-web password monitoring and quiet team phishing tests. Free Community tier; Pro ~$15/user/mo; 14-day trial, no card. Honest catch: it's MSP-first, and no email tool stops a phone call — the inbox is software's job, the phone is yours.The Operator's Move — "Verify before you pay," said out loud — Ten minutes, zero dollars. One rule: any request to move money or change an account gets confirmed by calling the person on a number you already had — never the number in the message. Add a money code word. Then tell your team out loud they're allowed to slow you down. It's the one defense that doesn't care how good the fake is.🔗 Show Notes & SourcesThe scam got a copywriter (AI-written BEC / FBI IC3)FBI 2025 Internet Crime Report — $20B+ fraud; 22K+ AI-related complaints, $893M+ (via KnowBe4): https://blog.knowbe4.com/fbi-report-americans-lost-20-billion-to-fraud-2025Business email compromise statistics 2026 — $3.05B / 24,768 cases / ~$123K avg; ~40% of BEC emails AI-generated (Hoxhunt): https://hoxhunt.com/blog/business-email-compromise-statisticsAI is making BEC nearly impossible to spot (ZeroHedge): https://www.zerohedge.com/markets/ai-making-business-email-compromise-nearly-impossible-spotEmail security statistics 2026 — native filters miss the AI-written 7%; AI detection catches 48% more BEC (Medha Cloud): https://medhacloud.com/blog/email-security-statistics-2026Three seconds of audio (voice cloning / deepfakes)Deepfake statistics 2026 — ~3 sec to clone a voice; 62% of companies hit; avg loss >$500K (Keepnet): https://keepnetlabs.com/blog/deepfake-statistics-and-trendsVoice cloning is the new BEC — deepfake CEO fraud (CybelAngel): https://cybelangel.com/blog/deepfake-ceo-fraud-how-voice-cloning-targets-us-executives/Your next "urgent call from the CEO" might be synthetic (Shumaker, Loop & Kendrick): https://www.shumaker.com/insight/why-your-next-urgent-call-from-the-ceo-might-be-synthetic-and-what-to-do-about-it/