# Fri Nov 6 · AI Update LIVE · 20:30 IST (10:00 AM ET) · 45 min

Theme: **"Your AI can spend, sell and get quoted. Who's checking it?"**
Built from this week's two clips (Nov 3 + Nov 4). On Thu night, swap in anything bigger that lands and drop the weakest segment.
Decks to screen-share: https://ai-update-2026-11-03.funneling.ai · https://ai-update-2026-11-04.funneling.ai
Rules: the news is the cold open, no intro about yourself, no proof lines. One CTA, once, at the end.

## Run of show
| Time | Segment | What happens on screen | Chat prompt |
|---|---|---|---|
| 0:00–3:00 | **Cold open** | "In a lab test, 9 of 10 AI agents gave themselves a bigger budget. One refused." Show the Nov 3 open slide. | "Has an AI tool ever done something you didn't ask it to? Type YES or NO." |
| 3:00–5:00 | **Rundown** | The 5 things we'll cover. Holding slide while people join. | "Where are you watching from?" |
| 5:00–13:00 | **1. Agents with a wallet** (Cornell Tech/Yale test) | Walk the test: approval codes in the inbox, the malicious email, the one model that refused. Hype check: lab test, no customer lost money. **Live demo:** open one of your own agent prompts and add the "never pay, refund or unlock without a human OK" line. | "Does any AI tool in your business touch payments or refunds?" |
| 13:00–21:00 | **2. Clicks are leaving Google** (HubSpot −80% traffic, revenue still up; DMG CTR drop; Google AI Mode ads) | Show the HubSpot story and Google's AI Mode ads post. **Live demo:** ask ChatGPT and Gemini "best [your niche] coach for X" and see whether you get named. | "Type your niche and I'll run the search live." |
| 21:00–28:00 | **3. What AI really costs** (Anthropic Q3 S-1 numbers, ChatGPT + Codex shared limit from December) | Hype check: the S-1 doesn't change your price today. The December shared limit does change how fast a Plus or Business seat runs out. **Live:** lay out one line per workflow: tool, monthly cost, what it replaced. | "How much do you spend on AI tools a month? Under $100 / $100–500 / $500+" |
| 28:00–33:00 | **4. Reality check: GDPx 12%** | The best model scores 12 out of 100 on hard, real-world work. So AI does the step and a human owns the outcome. | "Which task did you hand AI that it got wrong?" |
| 33:00–37:00 | **Security minute** | Stolen AI keys running up bills (Northflank/Wiz), the ChatGPT share-link flaw, VOLTA ranking victims with AI. One checklist: rotate keys, set spend caps, don't open strange share links. | "Done any of these this month?" |
| 37:00–43:00 | **Q&A** | Take the top chat questions. Answer each with one move they can make today. | — |
| 43:00–45:00 | **The one CTA** | "If you want to watch an installed AI system run, live: the free weekly session is at **ce2-live.funneling.ai**. Next one is Wednesday." Put the link in chat once. | — |

## Prep checklist (Thu night)
- [ ] Re-scan 24 h of news, swap a segment if something bigger landed (Claude can run `/ai-update` Thu evening)
- [ ] Open both decks, press **P** for presenter notes
- [ ] Have one real agent prompt ready for the segment-1 demo, with no client data on screen
- [ ] Have ChatGPT + Gemini tabs open for the segment-2 search
- [ ] Have someone (Leela) pin the session link ready to post at 43:00, and clip 3 moments afterwards for Shorts

[NIKHIL: one 30-second story from your own week where an AI tool surprised you. It goes in segment 1 or 4.]