I remembered that on one of our accountability calls last week, somebody mentioned a restaurant owner who was interested in an AI receptionist but wanted to know whether it could help filter out all of the spam calls restaurants get. At the time, there really wasn’t a clean answer for doing that directly inside the Voice AI agent. Well... apparently somebody at GoHighLevel was listening. 😂 They just released a new feature called Smart Call Ending & Spam Caller Protection for Voice AI. From what I’m seeing, the Voice AI agent can now: - Detect telemarketers/spam during the actual call and end the conversation - Analyze completed inbound calls for spam - Automatically block repeat spam callers by placing them on inbound DND - Let you choose how many spam calls have to happen before a caller is blocked - Notify you when a spam caller has been blocked - End legitimate calls more naturally instead of just awkwardly cutting off So for whoever had the restaurant prospect last week, this might be something worth going back to them with. If you’re using Growth Hub 365, the feature should be under: Settings → Labs → Voice AI: End Call & Spam Protection Then once it’s activated, the controls are available inside the Voice AI agent. One thing to know: HighLevel says the newer Speech-to-Speech agents don’t support the spam-protection feature yet, but support is supposed to be added. And speaking of Speech-to-Speech... There’s another Voice AI update that might be worth testing. HighLevel now has newer real-time Speech-to-Speech models that are designed to make the conversation sound more natural and reduce some of that noticeable pause between the caller speaking and the AI responding. There are currently two options I’m seeing: Gemini RealTime — $0.10/minute GPT RealTime — $0.20/minute So if anybody has been testing Voice AI and felt like the pauses made the agent sound a little too “AI-ish,” these newer models may be worth comparing against your current setup. Just thought I’d share because the spam-call question immediately made me think about that restaurant owner from last week.

which one? https://chatgpt.com/s/m_6a8a2898eb9c8191a0d4eb8be696084f

I originally started building my Front Desk Revenue Leak Audit in AI Studio, but I ran into CRM field-mapping and identity issues, so I rebuilt the core system natively in GoHighLevel. Right now, the working flow is: GHL Survey v2 → contact created in CRM → 8-question audit → scoring → tier-specific result page → completion workflow All three result paths have passed end-to-end testing: - Strong Front Desk - Some Revenue Leaks - Significant Revenue Leaks The survey is creating the contact, saving the audit answers, calculating the score, routing to the right result page, and triggering the workflow. So I feel good about the actual Lead Capture Machine. What I’m focused on now is the next part: what happens after somebody completes the audit but doesn’t take another action. Based on Monday’s conversation about follow-up being a decision system and not just an email sequence, I’m thinking about that behavior like this: Someone completing the audit shows interest, but I don’t want to automatically assume they’re ready to buy or book a demo. So I’m trying to determine: 1. What should the first follow-up be after they complete the audit but take no further action? 2. How long should I wait before sending it? 3. Should that first follow-up be automated, AI-assisted, human, or a mix? 4. At what point should the follow-up become more direct if they click the Digital Front Desk demo, visit the booking page, or start booking and don’t finish? I’m also still finishing some polish on the audit experience and the post-audit CTA, but I’m trying not to get stuck rebuilding instead of actually using the machine. If anybody wants to run through the lead magnet and give me feedback, you can try the Front Desk Revenue Leak Audit here: https://audit.thewheelchairpreneur.com I’d especially love feedback on the experience itself and how you all would handle that first “audit completed, no next action” follow-up. Side note: I put this update together with ChatGPT using the memory and context from all the work we’ve been doing on the Digital Front Desk build. That helped me pull the whole picture together without having to remember every little detail myself — which actually ties right into Monday’s conversation about giving AI better context instead of starting from scratch every time.