The Workshop Was a Fire Hose (In the Best Way)

This weekend we ran our first-ever Email Deliverability Workshop… and yeah—people showed up.

Big time.

The vibe was basically: 4 hours of dense, tactical “ohhh THAT’S why my emails go to spam” type of learning. No fluff. No motivational speeches. Just real infrastructure, real fixes, and real answers.

A bunch of people told us the same thing afterward:

• “It felt like drinking from a fire hose.”

• “So many golden nuggets.”

• “I’ve got a checklist for the next 30 days now.”

• “I’m actually confident I can fix inbox placement.”

Which is the entire point.

Because if your emails aren’t landing where humans can see them… your funnel is basically yelling into a pillow.

What We Covered (The Stuff That Actually Moves Inbox Placement)

We walked through the things that actually control whether Gmail/Yahoo/Microsoft treat you like a legit sender—or a suspicious raccoon in a trench coat:

• Inbox placement fundamentals (and why “delivered” is not the same as “seen”)

• SPF / DKIM / DMARC setup without needing a network engineering degree

• Domain alignment + infrastructure strategy so your marketing emails don’t torch your main domain

• Warmup + scaling frameworks so you can increase volume without getting throttled

• Monitoring setups (Postmaster Tools, SNDS, Yahoo tools) so you’re not flying blind

• Engagement protection systems so your list doesn’t rot quietly over time

VIP Q&A Was the Secret Sauce

Instead of a basic one-hour Q&A, the VIPs got a two-hour live Q&A—and it turned into one of those rare sessions where everyone’s problem is also your problem, so you’re taking notes even when it’s not your question.

We hit everything from:

• “Why am I in Promotions?”

• “Do subdomains actually matter?”

• “What do I do if I’m already in spam?”

• “How do I scale without blowing up reputation?”

…and we stayed until it was handled.

Want the Replay + VIP Q&A?

If you missed it live but still want to get access, you can grab the recordings here (this includes both sessions + VIP Q&A + bonuses):

https://live.automatedmarketer.net/vip-after-event

💰 The Offer Comes First (Or Nothing Else Matters)

You can build the cleanest snapshot, the prettiest funnel, the most “automated” automation that ever automated… and still make exactly $0 if your offer is mushy.

That’s the heartbeat of this Start From Scratch chapter: don’t start building the agency… start building the offer. Because if you don’t know what you sell (and who it’s for), you’re basically assembling IKEA furniture without the box photo. Technically possible. Emotionally catastrophic.

Here’s the thing… most people don’t fail because they can’t do HighLevel. They fail because they walk into the market with something like:

Cool. So… nothing. That’s not an offer. That’s a cry for help.

What they’re pushing here is a way tighter approach: pick a niche that’s already spending money, already has leads, already understands sales — and then fix the leak that’s quietly murdering their growth.

That’s why they land on appointment-based businesses (chiro, dental, med spa, home services, etc.). These businesses have one beautiful trait:

They can usually tell you what an appointment is worth.

Meaning you can charge real money… because the ROI math is obvious.

The core idea: fix the “leaky bucket”

A lot of these businesses already have leads coming in (ads, organic, referrals), but the follow-up is sloppy:

• missed calls

• after-hours leads going cold

• slow responses

• no-shows

• inconsistent reminders

• “we’ll call you back” (narrator voice: they never call you back)

So instead of promising the world, the play is:

Turn more of their existing leads into booked appointments — using a mix of Voice AI, Chat AI, and automation sequences.

The 4-core-tools offer (simple on purpose)

Even though HighLevel can do 400 things, the recommended “starter agency” stack is basically:

• Voice AI (inbound + eventually outbound with rules)

  • Handles missed calls, after-hours calls, busy staff moments

  • Transfers to humans when needed

• Chat AI

  • Website chat widget that qualifies and pushes bookings

  • Stops “dead website syndrome” (traffic with zero action)

• Automations

  • reminders, follow-ups, reactivation, no-show recovery

  • solves the real appointment killer: people not showing up

• Website / landing page (only if needed)

  • if their site is weak, you give them a clean booking path

  • doesn’t need to be fancy, it needs to convert

So the offer becomes:
“We install a system that answers more leads and turns more of them into booked appointments — 24/7.”

Not “we do marketing.”
Not “we do AI.”
Not “we build funnels.”

Appointments. Period.

The framework that makes your pitch land in 10 seconds

They call it the 4 Ps:

• Person (who it’s for)

• Problem (what’s broken)

• Product (what you install)

• Payoff (what they get)

Example structure (their vibe, simplified):

If they’re not that person? Perfect. You stop talking and move on.
This is volume. Not convincing. Not begging. Not “let me customize 10 things for you.”

Risk reversal: the “no-brainer” that gets you your first clients

This part matters a lot if you’re new and have no case studies.

Their angle is basically:

• Make the offer so safe it feels stupid to say no

• Use a clear, measurable guarantee

• Add reasonable conditions so you don’t accidentally guarantee miracles to businesses with zero leads

Examples they discuss:

• “30% more appointments in 30 days or you don’t pay / money back”

• Condition it on things like:

  • minimum lead volume

  • the business actually responding when needed

  • there being room to improve (if they’re already converting 9/10 leads, don’t promise a lift)

And there’s a sneaky psychological win here:

If you miss the guarantee slightly and lead with, “I’m refunding you,” a lot of clients go:

That’s how risk reversal builds trust fast.

The retention “secret sauce” (and it’s not magic tech)

The line that hits hardest:

“Make them feel like you care more about their business than they do.”

People will forgive imperfections if:

• you communicate

• you take ownership

• you’re proactive

• you’re clearly not hiding

It’s the same reason nobody fires the cleaner who genuinely cares (even if they keep putting the cups in the wrong cabinet like it’s an escape room).

💡Pro Tip (I learned this the hard way):
When you write your offer, force yourself to say it out loud in 10 seconds without stumbling. If you can’t say it cleanly, your prospects definitely can’t understand it cleanly. And if they can’t repeat it to their business partner, you’re not getting the sale.

You don’t need a better funnel.
You need a better sentence.

And once that sentence is tight — then you build the system.

Because building without an offer isn’t “being proactive.”
It’s just procrastination wearing a HighLevel hoodie.

🚨Claude Code Is the New “Learn Funnels”… Except It Actually Matters

I really want to make something clear.

I’m going all-in on Claude Code.

Not casually experimenting. Not “playing with it on weekends.” I mean actually investing time, building with it, understanding it properly.

Because the future is not more prompts.

It’s AI agents. Agent deployment. Automations. Systems that run without you babysitting them.

Very soon, AI agents won’t just “assist” your business — they’ll be running large parts of it. Follow-up. Development. Optimization. Deployment. Internal workflows. Product builds.

If you’re not learning how agents are structured and deployed now, you’re going to feel behind very quickly.

Claude Code is one of the cleanest bridges into that world.

And here’s what nobody’s telling you…

Claude Code Is the New “Learn Funnels”… Except It Actually Matters

The install is the least important part.

You can install it in 5 minutes, feel like a wizard for 48 hours… and then three days later you’re staring at your screen like:

That’s not because Claude Code isn’t good.
It’s because most people are doing the steps in the wrong order.

And the order matters a lot.

The future is not “better prompts.”
The future is AI agents + deployment + automations running the boring (and increasingly the important) parts of your business.

Claude Code is basically the bridge between “AI is cool” and “AI actually ships.”

What Claude Code Actually Is (In Normal-Human Terms)

Think of Claude Code like a developer who:

• plans the build

• creates and edits files in the right places

• keeps the project organized

• doesn’t disappear right when deployment gets messy

It’s not like most “build me an app” tools that generate a pretty front end and then leave you stranded when it’s time to actually deploy.

Claude Code is built for real application development.

But plot twist: you can make it beginner-friendly… if you set it up correctly.

The Trap Everyone Falls Into

Most tutorials teach this:

1. Install Claude Code

2. Start building immediately

3. Panic later

4. Burn credits

5. Blame the tool

The cheat code?

Build last. Not first.

Because if you start building before guardrails exist, Claude starts freelancing inside your repo like a chaotic intern with admin privileges.

The Correct Order (That Saves Credits and Sanity)

If you want this to feel like leverage instead of chaos, the order looks like this:

1. Install Your Foundation

• Node.js (package ecosystem)

• VS Code (your “house”)

• Claude Code extension

• Git Bash (Windows users)

• Workspace trust configured

None of this is sexy. All of it is necessary.

2. Create Guardrails BEFORE You Build

You need a claude.md file (both global and project-level).

This is your project constitution.

Without it, every build starts like Groundhog Day with amnesia.

No structure. No memory boundaries. No expectations.

That’s where credit burn starts.

3. Add Capabilities (Skills)

Claude can download “skills.”

Think of these like add-ons or specializations.

If Claude doesn’t know how to do something out of the box, you can install a skill and connect it to the relevant app.

Now it’s not just chatting. It’s executing.

4. Use Commands to Deploy Mini-Armies

This is where it gets interesting.

Instead of one giant agent doing everything (and rereading your entire project every time), you:

• deploy smaller specialized agents

• each with tighter context

• each handling one domain

• then reporting back to the main agent

This reduces drift.
This reduces context overload.
This reduces cost.

Why Agents Save You Credits (Non-Technical Version)

When you use one mega-agent, it keeps reloading your entire history and context.

More memory = more tokens = more cost.

When you use sub-agents:

• they operate in smaller scopes

• with smaller memory windows

• then return clean output

Less rereading.
Less confusion.
Less “why did it rewrite half my project?”

It’s the same reason you don’t have your CEO answering support tickets at midnight.

Specialization scales.

Why This Matters for Business (Not Just Coding)

Here’s the bigger picture.

Businesses are about to be run by:

• agent systems

• automated execution layers

• intelligent routing

• deployment frameworks

Claude Code is not “just a coding tool.”

It’s training you to think in:

• systems

• structure

• delegation to AI

• multi-agent orchestration

That’s the skill.

And that skill is going to matter way more than knowing how to build another landing page.

💡Pro Tip (Save Yourself a Ton of Credits)

Before Claude touches a single file, your claude.md should define:

1. Project structure expectations (where files go, naming conventions)

2. Build/run/test/deploy rules

3. Hard guardrails (no secrets in code, no random refactors, no invented dependencies)

If you skip this, Claude will still help.

But it’ll help like a toddler helping you cook.
Enthusiastic. Creative. And flour everywhere.

So yeah.

I’m going all in.

Because the next advantage isn’t “who has the best funnel.”

It’s who can spin up a reliable agent workflow that ships outcomes on command.

If you start practicing that now, you’re not lucky.

You’re early.

Claude Code SOPs (Installation + Setup + Templates):
https://drive.google.com/drive/folders/1CFfShM5NTeM4lzuMs7dQMay0OdpdzO3h?usp=sharing

The Week AI Quietly Moved From “Cool” to “Controlling Stuff”

AI news moved so fast this week it basically tripped over its own shoelaces. I’m not here to waste your time, so let’s go straight to what actually matters (and what you should steal for your business).

Nano Banana 2: “Free, Fast, and Weirdly Good at Text”

Google DeepMind dropped Nano Banana 2, and it’s basically Nano Banana Pro… just faster and more accessible. The practical headline is: you can use it for free in places you’re already in (Gemini / AI Studio / Flow / Vertex), and it’s good enough that you’ll stop saying “AI images always mess up text” like it’s still 2024.

What stood out:

• Speed toggle = real time saver (fast mode cuts generation time roughly in half)

• Search grounding: it can pull from real-world info + images via web search so you can generate specific subjects more accurately

• Text rendering + translation: fewer typos, fewer cursed letters, more “oh wow, that actually reads correctly”

If you want the simplest “try it now” workflow:

• Go to gemini.google.com

• Click Create an image

• Pick a style (you’ll see more style options now)

• Flip the toggle to Fast (that’s your “Nano Banana 2” tell)

• Generate something practical like:

  • a product mockup (enamel pin example)

  • an infographic (distillation explainer example)

The bigger point: image models are quietly becoming business tools, not toys. When the text stops breaking, these become usable for:

• quick landing page graphics

• ad creative variations

• lead magnets

• product mockups

• infographics you don’t have to “fix in Canva for 45 minutes”

💡Pro Tip: Don’t start with “make me a beautiful image.” Start with a format: “make an infographic,” “make a product mockup,” “make a YouTube thumbnail.” Models behave better when you give them a job title, not a vibe.

Perplexity Computer: The “Turnkey Agent” Era Is Here (And It’s Not Subtle)

The most important theme of this week wasn’t one shiny model—it was the fact we’re officially in the era of agents that can actually do work.

Perplexity introduced Perplexity Computer, and the pitch is basically:

It can research, design, code, deploy, and manage projects end-to-end in a safe cloud environment. And the key difference they’re pushing versus the more DIY agent world (like OpenClaw) is:

• Perplexity Computer = turnkey

  • they handle the environment

  • they handle the orchestration

  • they try to handle the “oh no it deleted my inbox” risk

• OpenClaw = autonomous + user-owned

  • you install locally or on your own VPS

  • you own the data

  • you also own the risk (including the “why is it speedrunning chaos?” moments)

Perplexity Computer’s flex is orchestration:

• routes work across 19 models

• uses different models for different tasks (reasoning, images, video, speed tasks, long context, etc.)

• spins up sub-agents to work in parallel (research while drafting, data gathering while building, etc.)

Downside (and it’s not small):

• Max-only right now ($200/mo / $2,000/yr tier)

Still, the demos show what’s coming:

• build an interactive S&P 500 bubble chart site (end-to-end)

• create animated stock visuals (the kind that literally goes viral)

• handle task lists by breaking them into subtasks and executing autonomously

And there was a perfect “this is why containment matters” moment:

• a safety researcher joked about telling an autonomous agent “confirm before acting”… and watching it attempt to delete their inbox anyway (and not stopping until they physically ran to their computer)

That story is funny until it’s your Stripe account.

💡Pro Tip: If you’re going to play with autonomous agents, give them “read-only” defaults and escalation rules. The first automation you let an agent do should never be “delete,” “send,” or “publish.” Make it summarize, draft, and propose. Then you approve. (You can still move fast without gambling your business.)

NotebookLM Guide: The Underrated “No Hallucinations” Productivity Weapon

This part matters because it’s actually usable for normal humans.

NotebookLM is still one of the most underrated free AI tools because it can take your sources (notes, docs, articles) and turn them into structured insight—without the usual “confident nonsense.”

The guide highlights:

• how to turn messy notes into structured takeaways

• how to ask better questions against vetted sources

• how to use AI like a research assistant instead of a fiction writer

If you’re buried in saved articles, research, SOPs, client notes—NotebookLM is the “stop doomscrolling and start extracting” tool.

Microsoft Copilot Tasks + Cursor Agents: Everyone Wants an Agent That Runs While You Sleep

Microsoft is working on Copilot Tasks (waitlist right now), and it’s basically their answer to the same trend:

• go from “drafts” → “completed tasks”

• recurring tasks (briefings, urgent email surfacing, draft replies)

• scheduling + execution inside the Microsoft ecosystem

Cursor also rolled out agents that can control their own computers (with a twist I actually love):

• it records what it did in the virtual machine

• you can review the “agent session” like a replay

• you can set it to run for hours (“3 hours / 5 hours / 10 hours / until done”)

This is the pattern:

• agents are becoming workers, not features

• multi-model switching is becoming standard (use one model until it stalls, switch when it gets stuck)

QuiverAI: SVG Art From Code (And It’s Weirdly Satisfying)

Quiver generates SVG images (code-based visuals) instead of diffusion-style images. It’s slower, but watching it “draw” live is oddly addictive.

Why it matters:

• SVGs can be copied into websites

• potentially animated

• scalable and lightweight

• useful for web assets, icons, graphics without giant image files

Not perfect, but very fun—and surprisingly usable for web stuff.

The Anthropic vs Pentagon Drama: This Is the One That Actually Has Teeth

This is the heavy one.

The Pentagon wants Anthropic to loosen two safeguards:

• no mass domestic surveillance

• no fully autonomous weapons without human involvement

Anthropic’s position is basically:

• lawful foreign intelligence and counterintelligence: fine

• mass domestic surveillance: incompatible with democratic values

• fully autonomous weapons: models aren’t reliable enough and shouldn’t be used that way yet

The Pentagon threatened:

• “supply chain risk” designation (a massive deal)

• potentially invoking the Defense Production Act to force safeguards removed

And the contradiction is wild:

• “You’re a security risk” vs “You’re essential to national security”

Meanwhile, the pressure valve is obvious:

• xAI (Grok) reportedly agreed to the “all lawful use” standard for classified systems
  So even if Anthropic stands firm, the government can shop elsewhere.

This isn’t “AI gossip.” This is precedent-setting stuff.

💡Pro Tip: If you run anything regulated, client-sensitive, or privacy-heavy—start assuming AI vendors will get pulled in different directions by governments and policy shifts. Build with portability in mind (so you’re not married to one provider forever).

Distillation Attacks + The Irony Nobody Can Ignore

Anthropic also published on detecting/preventing distillation attacks, because Chinese companies allegedly used Claude outputs to distill competing models.

But the internet’s reaction is… predictable:

• “You scraped the internet without permission. Now you’re mad someone scraped you.”

The real takeaway isn’t the meme—it’s that model extraction, distillation, and imitation are going to keep accelerating. The moat is shifting from “model weights” to:

• distribution

• product experience

• proprietary data

• agent workflows

• integrations

Rapid Fire: Weird, Wild, and Slightly Dystopian

• Claude scheduled tasks (recurring work at set times)

• Claude Code desktop preview for running apps (quality-of-life win)

• Claude remote control: approve actions from your phone so the agent doesn’t stall waiting for you

• Notion custom agents across Notion/Slack/Mail/Calendar/Figma/Linear/MCP servers

• Standard Intelligence claims a general computer action model trained on 11M hours of video (this one could be a sleeper story)

• Pika AI Selves: like a Tamagotchi with memory and social media ambitions (cool… but unclear use case)

• Sam Altman comparing training AI to raising humans (a take that made a lot of people stare into the middle distance)

• OpenAI device rumors: this month it’s “smart speaker”

• Galaxy S26 event: more AI in everything (try-ons, scam detection, etc.)

• Burger King using AI to check if employees say please/thank you (welcome to the future, I hate it here

Nano Banana 2 Is Basically “Pro”… But on Fast Forward

Nano Banana 2 (aka Gemini 3.1 Flash Image) is Google’s new image model, and the pitch is simple: Pro-level quality at flash speed—and you can use it free in a bunch of Google surfaces.

And annoyingly… it mostly delivers.

Here’s what matters:

• Speed is the real upgrade

  • Fast mode regularly lands around 13–15 seconds

  • Pro is often 25–35 seconds

  • Quality difference is there, but for most stuff it’s close enough.

• Text rendering is finally usable

  • It handled a brutal “pricing page UI” prompt with headlines, tables, footnotes, exact labels… and didn’t turn the text into spaghetti.

• Translation inside images is solid

  • It translated an event poster to Spanish while keeping layout consistent.

  • Fast even beat Pro on one translation test.

• Consistency is strong… until you change camera angles

  • It can keep characters + objects consistent across scene edits.

  • But when you ask for a new camera angle, it can reshuffle the room (and occasionally delete a character like it never existed).

• 4K output is still a question mark

  • It claims 4K, but downloads often came out at 2752 × 1536 even when asked for 4K.

  • Still looks sharp enough for most marketing use.

How to try it fast:

• Go to gemini.google.com → Create an image

• You’ll know it’s Nano Banana 2 because you’ll see more style templates

• Flip to Fast and generate

Nano Banana Pro is now mostly paid, but you can still generate with 2 and hit Redo with Pro if you’re on a paid plan.

Bottom line: for like 95% of marketing use cases, Nano Banana 2 is going to be the daily driver—fast, clean, and finally reliable with text (which is the part that actually matters).

💲Deals at a Glance

Our GoHighLevel Affiliate Offer
Best Affiliate Offer For GoHighlevel On The Planet.
Get Started Now

🔥 Vibe Coding Incubator –
Get hands-on with Vibe Coding, live weekly trainings, app templates, support, and Lovable Ai How To’s
👉 Join For Free

😲Best Gohighlevel Course For Only $20!
1000+ Student 4.7 in reviews and over 18 hours of content: https://www.udemy.com/course/gohighlevel-masterclass-from-beginner-to-expert/?couponCode=GHLSPECIAL

Join Our Free Skool Community and Course
Build your agency or side hustle with HighLevel, n8n, AI & Automation. Your path to quitting the 9-5 starts here.
Join Now

Marketing Tech Professionals Ready For Hire (Done For You Work)
MarTech professionals trained by me, ready to join your team for any marketing technology needs.
Explore More

VIP Office Hours
For those who just want simple questions answered by real knowledgeable human beings. Join our group office hours 4 days a week.
Access VIP Office Hours

Remember, there’s always something new coming from the Automated Marketer YouTube Channel—new content drops every Tuesday & Thursday. Subscribe so you never miss the next big idea.

Keep watching, keep learning, and keep moving forward!!

Nuno