I build AI systems that automate your lead generation | Founder @ Leva AI | Amazon, ex-P&G, ex-PwC

Most people trying to start an AI Automation Agency get stuck.

They jump between YouTube videos, random tutorials, and overpriced communities… and still:

- Don’t know what to build
- Don’t know how to get clientsù
- Don’t feel technical enough
- Don’t have a clear plan

I’ve been there.

A few years ago I knew nothing about AI, nothing about automations, nothing about landing clients. I wasn't technical at all. I was working a 9-5, trying to learn this stuff alone, and wasting months on trial and error.

Then I built my first automation.
Then I landed my first client.
Then I made my first $10K.

And 5 months after: I made more than $50k, and grew my LinkedIn followers to 30k+.
And after that, I managed to leave my 6 figures salary, and start chasing the life I always wanted.

And now, mentor individuals who want to do the same.

Because I know what it feels like to have a vision but be stuck in a job you don’t like. I know what it feels like to be insecure, open n8n, and get completely overwhelmed. And that's why I can help you.

👉 Join other founders in my free Skool community here - https://www.skool.com/ai-automation-founders

Claude Code is one of the most powerful AI today.

(Especially after the Opus 4.6 update.)

Below the breakdown on to use it properly:

And if you want to learn how to create agentic workflows, and make money with it, check this: https://lnkd.in/d5nE_Fuc

1️⃣ What Claude Code really is

It’s not a chatbot. It’s an AI that can work inside your computer.

↳ It can read your files
↳ It can edit documents
↳ It can organize folders
↳ It can run tasks automatically
↳ It can handle long, complex projects without forgetting

Chat Claude (browser version) = answers questions.
Claude Code = completes real work.

2️⃣ What it can actually do (in simple terms)

Imagine giving instructions to a very fast digital assistant. Claude Code can:

↳ Turn messy meeting notes into a clear business document
↳ Read thousands of files and summarize them
↳ Create reports automatically
↳ Organize your project structure
↳ Connect your tools together

Instead of switching between 10 apps, it works in one place.

You describe the outcome.
It executes the steps.

3️⃣ The workflow that makes it powerful

This is the simple pattern:
Understand → Plan → Build → Improve

Example: Meeting recording

→ Clear plan document
→ Task list
→ Summary message for the team
→ Progress overview

All generated step by step.

4️⃣ How to get started

You don’t need to be an engineer.
You need:

↳ An Anthropic account
↳ A Pro or API plan

Then:

Open your computer’s terminal
Type “claude”
Press Enter

You are now inside your AI workspace.
Anthropic provides official step-by-step setup guides.

5️⃣ Connecting your tools (MCP)

This sounds technical, but it’s simple.

MCP is just a way to let Claude talk to your other apps. Think of it like a universal adapter.

It allows Claude to connect with:

↳ Google Docs
↳ Slack
↳ Notion
↳ Email
↳ Databases

So instead of copying and pasting between apps, Claude moves information for you.

6️⃣ Important commands (basic control)

These are just shortcuts:

/help → see what Claude can do
/clear → start fresh
/model → switch AI version
/mcp → check connected apps

You don’t need to memorize everything.

7️⃣ How to make it smarter over time

Two simple ideas:

↳ Reusable instructions (called “Skills”)
↳ A project memory file (called CLAUDE.md)

In simple words:

You can save your preferred style, rules, and structure
So Claude remembers how you like things done.

It becomes consistent.

8️⃣ The right way to give instructions

Most people type one sentence and hope for magic.

Power users do this instead:

Step 1 → Ask it to create a plan
Step 2 → Review the plan
Step 3 → Approve it
Step 4 → Let it execute

If something goes wrong:

Stop
Adjust
Continue

Claude Code isn’t powerful because it “talks well.”
It’s powerful because it works.

And once you understand that difference, everything changes.

♻️ Repost to help your network

Giovanni Beggiato

I help founders scale to $10K/mo+ with their AI Automation agencies, from zero | Made $100k+ in 5 months with mine | Join other AI Agency owners in my Skool community (Link in the featured section)

Claude Cowork is a non-negotiable for AI Agency Owners.

Here's the full breakdown so you can get started: 

Most people think AI assistants just answer questions.

Cowork actually does the work.

Files. Browser. Emails. Apps.
All from your computer.

No code. No developer. No waiting.

Here's the full breakdown:

🔸 What Cowork actually is

It's Claude, but it lives on your desktop.

It doesn't just respond.
It acts.

↳ Opens and edits your local files
↳ Controls your browser
↳ Runs multiple tasks at the same time (sub-agents)
↳ Builds real outputs: docs, decks, PDFs
↳ Connects to your tools via MCP and plugins

Built for founders, managers, and non-technical people.
(Claude Code is for developers. Cowork is for everyone else.)

🔸 How to get started in 4 steps

1️⃣ Download the Cowork app
2️⃣ Install the Chrome extension
3️⃣ Mount the folders you want it to access
4️⃣ Start simple — one task, one clear instruction

Don't add skills or plugins on day one.
Start vanilla. Add power only when you hit friction.

🔸 Real example: receipts → expense report in 3 minutes

↳ Scan your receipts
↳ Cowork organizes them by category
↳ Builds the spreadsheet
↳ Drafts the email to your accountant

That used to take 45 minutes.
Now it's 3.

🔸 10 simple prompts to get you started today:

1️⃣ "Clean up my desktop and organize files by project and date."

2️⃣ "Research [company name] and prepare a 1-page brief before my call."

3️⃣ "Scan these receipts and build a formatted expense report."

4️⃣ "Find the 5 most relevant articles on [topic] and summarize each in 3 bullets."

5️⃣ "Pull my notes from last quarter and write a Q3 review doc."

6️⃣ "Turn this one-pager into a 10-slide pitch deck outline."

7️⃣ "Take this video transcript and repurpose it into a LinkedIn post, email, and summary doc."

8️⃣ "Transcribe my meeting notes and extract action items by owner."

9️⃣ "Sort my inbox by priority and draft responses to anything urgent."

🔟 "Run these 3 research tasks in parallel and bring me a combined report."

The difference between learning AI and using AI
is prompts like these.

One saves you time.
The other just looks good on LinkedIn.

If you want to go further and actually build and sell AI services,
check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

33 pages. 5 chapters. 
The Complete FREE Guide for AI automation.

Everyone wants to get into AI automation, 
but nobody explains the basics properly.

People are thrown into tools
↳ without a map
↳ without context
↳ without understanding what they’re doing

So I made the guide I wish I had when I started.

A guide for beginners entering the space, 
and for anyone serious about AI.

Here’s what’s inside:

1️⃣ What automation actually is

↳ Rule-based vs AI-powered
↳ What makes a task automatable
↳ Triggers, actions, conditions
↳ How to evaluate ROI before building

2️⃣ How n8n really works

↳ Triggers vs logic vs actions
↳ How nodes think
↳ How data moves between steps
↳ Cloud vs self-hosting, and why it matters

3️⃣ JSON, expressions & data flow

↳ Arrays, items, nested fields
↳ How every node outputs data
↳ How to reference previous nodes
↳ How expressions actually work

This is the part most beginners never learn, and the reason their workflows break.

4️⃣ Workflow design fundamentals

↳ Map the process before building
↳ Validate inputs early
↳ Clean structure, clean naming
↳ Test in small chunks
↳ Error branches & recovery workflows
↳ How to keep systems reliable at scale

This is where you shift from “playing with automations” to building real systems.

5️⃣ AI agents inside workflows

↳ What an agent is
↳ The 4 agent types
↳ When to use an agent vs a simple LLM call
↳ How n8n integrates AI
↳ How to prompt agents for consistent behaviour
↳ How to structure multi-step reasoning

This is how you add intelligence to your automations, not just speed.

🔸 Why this guide matters

It gives you the foundations of:
↳ How automation works
↳ How tools think
↳ How data moves
↳ How AI fits into real systems

If you want to learn AI automation properly, this is where you start.

And here you can find the video walkthrough I made for this:
🔗 https://lnkd.in/d5nE_Fuc

The AI Automation Started Guide (2026 Edition)

CLAUDE.md is a 1-page file that will 10x your AI output...

or destroy your Agentic Workflow ⬇️ 

This is based on internal best practices shared by Boris Cherny, creator of Claude Code at Anthropic.

It's an operating system layer for how you work with AI, so that you can build anything, anytime.

But if you're also interested in selling AI services, 
you should check this out: https://lnkd.in/d5nE_Fuc

What's inside?

✅ Workflow orchestration
✅ Sub-agent delegation strategies
✅ Self-improvement loops
✅ Autonomous bug fixing
✅ Core principles for AI-led dev

Why it works?

It creates a feedback loop.

Instead of correcting the same AI mistakes over and over, this system captures your feedback as "rules."

The more you use it, the smarter it gets.
Stop fighting your LLM and start orchestrating it.

Comment CLAUDE, and I'll send you the full link.

♻️ Repost to help your network

undefined undefined

Stanford Online

Advanced Learning Algorithms

Supervised Machine Learning: Regression and Classification 

Unsupervised Learning, Recommenders, Reinforcement Learning

Imperial College London

Mathematics for Machine Learning: Linear Algebra

Mathematics for Machine Learning: Multivariate Calculus

Mathematics for Machine Learning: PCA

Python for Data Science, AI & Development

Udemy

100 Days of Code: The Complete Python Pro Bootcamp

University of Michigan

Python Data Structures

Programming for Everybody (Getting Started with Python)

Google

Google Digital Garage: Fundamentals of Digital Marketing

SDA Bocconi

Università degli Studi di Padova

Leva AI

Amazon

Procter & Gamble

Danone

PwC Italy

Precision Group

I build AI Agents for SMBs | Book FREE discovery call | Founder @ Leva AI | Speaker | ex-P&G, ex-Amazon, ex-PwC

Most businesses know they should use AI, but few know how.

The result?

🔸 Teams are stuck doing repetitive tasks manually. 
🔸 Leads go cold before they’re touched. 
🔸 Scaling feels impossible without hiring more people. 
🔸 And promising AI tools end up gathering dust because no one knows how to stitch them together.

That’s what I’m solving with Leva AI.

🛠 It’s a new AI automation agency I’m currently building to help SMBs transition to an AI-first operating model, without the overwhelm.

From inbound lead capture to outbound outreach, from internal workflow automation to AI agent integration, we build systems that actually work, tailored to your real-world processes. 

Our scope covers:

🔸 Lead Generation & Capture: Automated forms, chatbots, and scraping tools that bring leads in 24/7.
🔸 Lead Qualification & Nurturing: AI workflows that sort, score, and follow up with leads so your team speaks only with sales-ready ones.
🔸 Outbound & Sales Enablement: LinkedIn DMs, email outreach, and WhatsApp follow-ups, fully automated, always on-brand.
🔸 Customer Support & Service: AI agents that handle common questions, book appointments, and resolve issues, instantly and without burnout.
🔸 Internal Operations & Admin: Automate your back office, from scheduling and reporting to task tracking and handoffs.
🔸 End-to-End Tool Integration: Connecting tools like Airtable, Notion, Slack, Gmail, CRMs, and more, so your entire stack works together.

Leva AI is still in its early days. I’m currently validating, testing, and building in public.

Led by my experience as a former FAANG AI Program Manager and operations lead, the mission is clear:

You bring the ambition, I’ll build the systems to match.

If you're curious, open to a pilot, or just want to jam on ideas, let’s talk.
You'll be in good hands.

Book a FREE Discovery Call

Get FREE access to all my AI resources

Claude Code is a non-negotiable for AI Agency Owners.

Here's the full breakdown so you can get started:

Most people use AI like a chat window.
Claude Code actually does the work.

Terminal. Files. Code. Entire workflows.
All from your computer.

No switching tools. No copy-pasting. No dev team needed.

🔸 What Claude Code actually is

It's a command-line tool built by Anthropic.
It sits inside your terminal.
It doesn't just respond.
It executes.

↳ Reads and writes files directly on your machine
↳ Breaks your goal into steps and runs them end-to-end
↳ Connects to GitHub, Notion, Slack, Jira via MCP
↳ Remembers your project context via CLAUDE.md
↳ Builds reusable automations via Skills

Built for founders who want to build and ship fast.
(ChatGPT answers questions. Claude Code gets things done)

🔸 How to get started in 4 steps

1️⃣ Install Claude Code via your terminal
2️⃣ Open it inside your project folder
3️⃣ Hit Shift+Tab twice to activate Plan Mode
4️⃣ Write your goal clearly and let it map the steps

Don't set up MCP or Skills on day one.
Start vanilla. Add power only when you hit friction.

🔸 Real example: raw data → client report in 5 minutes

↳ Drop in your survey responses
↳ Claude Code categorizes them by theme
↳ Extracts key insights per category
↳ Writes the formatted report
↳ Ready to send

That used to take half a day.
Now it's 5 minutes.

🔸 10 prompts to get you started today:

1️⃣ "Read all files in this folder and give me a project summary."
2️⃣ "Analyze this survey data and categorize responses by theme."
3️⃣ "Turn these notes into a structured PRD."
4️⃣ "Build a dashboard from this CSV file."
5️⃣ "Research these 3 competitors and write a comparison report."
6️⃣ "Turn this transcript into a LinkedIn post, email, and summary doc."
7️⃣ "Write a CLAUDE.md for this project based on the files here."
8️⃣ "Clean up this dataset and flag any inconsistencies."
9️⃣ "Generate a pitch deck outline from this one-pager."
🔟 "Run these 3 tasks in parallel and give me a combined output."

The difference between learning Claude Code and using Claude Code is a plan.

One ships real work.
The other just looks good on LinkedIn.

If you want to go further and actually build and sell AI services, check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

I sold AI Agents to 50+ companies

Below is the material you need to do it to

In order, below you'll find:
↳ Platforms to use (my personal reco)
↳ Theoretical Stuff (Repo, Books, Manuals)
↳ Business Part (how to scale an Agency to 10k/mo) 

🔸 Platforms to use:

🔗 Antigravity for Agentic Workflows
https://lnkd.in/dvPzca7Z

🔗 n8n for AI Automations
https://n8n.io/

🔸 Theory: How to build AI Agents (Repo, Books & Manuals)

🔗 10 GitHub Repos Every AI Agent Builder Must Know
https://lnkd.in/dDyXnUXq

🔗 AI Agents Are More Than n8n + ChatGPT (Here’s Proof)
https://lnkd.in/di6Qe_Y7

🔗 The Ultimate AI Agent Roadmap (Step by Step)
https://lnkd.in/de8ghChA

🔗 50M New AI Jobs Are Coming - Here’s How to Prepare
https://lnkd.in/dugbbfnP

🔗 What 200 Hours of AI Agent Study Taught Me
https://lnkd.in/dDYWmRay

🔗 10 AI Tools Every Founder Uses to Scale Smarter
https://lnkd.in/dA4ceRHt

🔗 OpenAI’s AI Agent Guide Explained in Plain English
https://lnkd.in/djpaFk6B

🔗 LangGraph vs LangChain vs Autogen vs CrewAI (Key Differences)
https://lnkd.in/dH8APX2x

🔸 Business Part: AI Automations

🔗 My Roadmap to scale to 10K/mo
https://lnkd.in/dYb4u6sE

🔗 Network of people building an AI Agency
https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

The one-person AI business one-pager.

6 steps to $100K with your AI Agency:

And if you want to build and scale yours, check this: https://lnkd.in/d5nE_Fuc

1️⃣ Spot the Pain
Don't start with tools. Start with problems.

Find something repetitive, costly, and ignored.

Then map it:
↳ Inputs → Steps → Outputs

That map is your product.

2️⃣ Start as a Service
Do the work manually first.

Prove people will pay before you automate anything.

Most people build before they sell.
That's why most people have zero clients.

Validation first.
Automation second.

3️⃣ Pick One Business Model

↳ Productized service: fixed scope, fixed price
↳ Cohorts/courses: time-boxed, outcome-driven
↳ SaaS/micro-tools: automate one specific workflow
↳ Templates/assets: passive sales

One model.
One market.
One offer.

Complexity kills momentum.

4️⃣ Build a Marketing Engine (4 moves)

→ Listen: find pain points via AI research
→ Create: build a site with Lovable + Searchable
→ Compound: get mentioned in ChatGPT and Perplexity
→ Personalise: tailor outreach with Clay + Mutiny

Most people post and hope.

This is a system.

5️⃣ Operations Rule

Simple decision tree:

↳ If it repeats → automate it
↳ If it needs judgment → templatize it

Tools that actually matter:
↳ Lindy → inbox and CRM
↳ Fathom → calls
↳ Cursor → automations
↳ Precision → KPIs

Your ops should run without you thinking about them.

6️⃣ Sales Engine

Capture → Qualify → Nurture → Close → Expand

AI handles follow-ups instantly.
AI revives no-shows.

Pitch deck structure that works:
↳ Problem
↳ Proof
↳ Plan
↳ Price

And the math is simple:

$3K × 3 clients = $100K/yr
$1K product × 9/mo = $100K/yr
$199 SaaS × 42 subs = $100K/yr

Pick your path.
Build the engine.
Repeat.

If you want to connect with others building and selling AI services,
check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

Voice AI Agents are going to be big in 2026.

I made a full course on how to build and sell them.

Voice Agent is where the real money is,
but nobody teaches how to build them properly.

So I recorded the guide I wish I had when I started.

And if you comment "VOICE",
I'll send you the link

Here's what's inside:

1️⃣ The Structure
↳ The 3-layer pipeline: Transcriber, Model, and Voice

2️⃣ The System Prompt
↳ The #1 mistake that causes hallucinations + how to write prompts that work

3️⃣ The Knowledge Base
↳ Why RAG beats stuffing info into the prompt

4️⃣ Tools
↳ 5 tool types to connect your agent to calendars, CRMs, Slack, and more

5️⃣ Customizing Your Voice
↳ 7 voice providers compare: quality vs speed vs latency

6️⃣ Latency Tuning
↳ Start Speaking & Stop Speaking Plans for real-time performance

7️⃣ Voice AI Agent Offers
↳ 5 business models ranked by income ($20K to $700K)

8️⃣ Voice AI Agents MVP
↳ The blueprint to build and ship each one

If you want to build and sell Voice AI agents,
this is the full playbook.

Comment "VOICE" and I'll send you the link

♻️ Repost to help your network

Most people use "AI Agent" to describe everything AI does.

That's a problem.

Because non-agentic AI, agentic AI, and AI agents are three completely different things.

and if you want to learn how to build and sell all three,
check this: https://lnkd.in/d5nE_Fuc

Here's the breakdown:

1️⃣ Non-Agentic AI
↳ One input. One output.
↳ No memory. No reasoning. No tools.

User types a prompt → model responds once → done.

No checking its own work.
No context retained after.

Use cases: summarizing reports, rewriting emails, generating headline variations.

Fast. Cheap. No setup.
But it can't handle multi-step work.

2️⃣ Agentic AI
↳ AI that plans, executes, critiques, and repeats on its own.

You set a broad goal → system breaks it into sub-tasks → connects to tools and APIs → evaluates results → loops to improve.

Use cases: research + market analysis + slide deck in one run, managing ad campaigns, generating and testing onboarding flows.

Handles complex projects.
Integrates with tools and memory.
More reliable than single prompts.

But it's slower, more expensive, and still needs human guardrails.

3️⃣ AI Agent
↳ A single-task worker that automates one repetitive job reliably.

You define one clear responsibility → agent connects to 1-2 tools (CRM, email, calendar) → executes → outputs result without human guidance.

Use cases: pull pipeline data from HubSpot weekly, draft FAQ replies and send via email, generate expense reports from accounting software.

Easy to test. Easy to refine. Reliable in its lane.

But it can't handle broader projects, and breaks with unclear inputs.

The confusion between these three is why people overpay for the wrong solution, or underbuild when the problem needs more.

Know what you're building before you build it.

If you want to learn how to build and sell all three,
check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

If you're running an AI agency in 2026... Claude Code is the one tool you need to master. This is how:

Most people are still using it like a chatbot. 
That's a mistake.

Claude Code is a full agentic system: 
↳ You assign a task
↳ It executes
↳ Uses tools

...and gets the job done.

But 99% of users don't know it has 12 features that completely change how powerful it is.. all in this repo:

1. Skills — reusable workflows you load on demand
2. Agents — custom sub-agents with their own tools, model, and permissions
3. Memory — persistent context across sessions via CLAUDE.md
4. Rules — modular instructions scoped by topic or path
5. Hooks — deterministic scripts that fire on specific events
6. MCP servers — direct connections to tools, databases, and APIs
7. Plugins — distributable bundles of skills, agents, hooks, and MCPs
8. Marketplaces — publish and discover plugin collections
9. Sandboxing — file/network isolation for safer execution
10. Output styles — configurable response formats
11 - Settings — 37 settings, 84 env vars, full control
12 - Permissions — fine-grained tool access with wildcard syntax

I also made this website, where I stored all best practices for you: https://lnkd.in/dDJTDRPV

The agency owners who learn this properly will build faster, deliver better, and charge more.

And if you want to start & scale an AI Agency,
check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

Most people building AI automations are failing.

Not because they can't build.

Because they can't deliver.
There's a massive difference.

Anyone can make an n8n workflow look impressive in a demo.
And in Antigravity? Even easier.

The real test is:

Does it still work 3 months after you handed it over?

For most agencies, the answer is no.

And here's why:

They never solved the fundamentals:

↳ Who owns the data and the account?
↳ Where do API keys actually live?
↳ What happens when something breaks at 2am?
↳ Who's responsible when scope creeps?
↳ What even counts as "delivered"?

Nobody teaches this part.

Everyone obsesses over the build.
The nodes. The prompts. The architecture.

But delivery is what keeps clients.
And clients are what keep businesses alive.

Here's the hard truth:

If you can't define Version 1 clearly...
If you can't separate delivery from upgrades...
If the client can't own and run the system without you...

You didn't deliver an automation.
You created a dependency.

That's not a service.
That's a liability.

The agencies scaling past $10K/mo aren't better builders.

They're better at handing things over.

Clean ownership.
Clear scope.
Real handover documentation.

That's what clients pay to keep.

If you want to learn how to actually deliver AI automations that stick, and connect with others doing the same, check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

How To Deliver Agentic Workflows (2026 Edition)

I quit my $150K job to open my AI Agency

Here are 3 things I understood right after:

1️⃣ Your salary never matches your value
And that’s not a criticism.
It’s the model.

As an employee, you’re paid to:
– fit a role
– follow a scope
– deliver predictably

Companies reward consistency, not upside.

As an entrepreneur, you’re paid for:
– absorbing risk
– dealing with uncertainty
– outcomes that may or may not work

More risk = more upside.

That’s why salaries cap early, even when impact doesn’t.

Same skills. Different rules.

2️⃣ You almost always quit too late
Never too early.

In a job, you’re trained to wait:
– until you "are ready"
– until you feel confident
– until everything is clear

That works in a 9–5.

Entrepreneurship works in reverse.

You don’t get clarity before acting.
You get it because you acted.

Waiting doesn’t reduce risk.
It delays learning.

3️⃣ Competence is not leverage
Outside a job, being good is assumed.

The market doesn’t ask:
“Who’s the smartest?”

It asks:
“Who can solve my problem, and do I know they exist?”

If people don’t know you:
– your CV doesn’t matter
– your background doesn’t matter
– your competence doesn’t convert

That’s why distribution matters.

No distribution → no leverage.
No leverage → no sales.

If you’re still in a 9–5, ask yourself:

– Is my upside capped by the model I’m in?
– Am I waiting for readiness instead of acting?
– Do I actually have a way to leverage what I know?

Different game.
Different rules.

Antigravity just changed AI automation forever.

And most people don't realize this: ⬇️

Before, building an AI automation required:
↳ Knowing how to code
↳ Understanding APIs
↳ Spending weeks on setup

Now?

You describe what you want.
It builds it.

That's it.

The barrier to build is collapsing.

Which means in 12 months, anyone will be able to create an automation with a simple prompt.

So if your edge is "I can build this"...

That edge is disappearing.

Fast.

But here's what doesn't change:

Most businesses don't need someone who can build.

They need someone who can answer:

↳ What problem does this actually solve?
↳ What is this worth to my business?
↳ How does this connect to real revenue?

The people who win won't be the best builders.

They'll be the ones who understand business problems well enough to position automation as a solution.

That's not a technical skill.
That's a business skill.

And business skills don't get automated.

The game isn't "who can build the fastest."

It's "who can walk into a room, identify the money leak, and show exactly how automation fixes it."

That's where the real margin is.

That's where clients stop comparing prices.

If you want to learn how to build and sell AI automations with real business value, check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

I spent $52K+ in coaching programs in 2025

Here's what I learned (and the outcome):

1️⃣ Not every investment returns what you expect
And that's okay.

The goal isn't to be perfectly right.
It's to be directionally right.

I paid $6K for a LinkedIn course.
Didn't monetize it right away.
But I built an audience.

Not a loss.
A foundation.

2️⃣ Technical skills don't make you an entrepreneur
I paid $6K for an AI course.

I became genuinely good at:
– building AI solutions
– understanding the tech
– closing deals on paper

Zero clients after.
ROI = 0.

Because there's a massive leap between:
– being able to build something
– being able to sell it

Competence is not a business.

3️⃣ The scariest investment was the best one
$36K+ in business coaching.

At the time, that was almost everything I had.

My coach told me something I still carry:

"If your salary pays your bills and puts food on the table... this money isn't for survival.

So the question isn't whether to invest it.
It's whether you trust yourself enough to."

You'll pay either way:
– with money → you buy time
– with time → you pay with mistakes

That $36K generated me $100K+ within 5 months.

Most people aren't afraid of investing money. 
They're afraid of betting on themselves.

But if you don't believe in yourself... why would anyone else?

Have faith in your next step. 
It's the only bet that truly matters.

---

PS: in case your are a young entrepreneurs, starting out and no money to invest, these are my 2 favourite books. Start there!

Complete Guide on How to Price Agentic Workflows

I made $60K+ with this pricing system:

Everyone wants to make money with AI workflows,
but pricing is the part nobody teaches you.

How do you define the value?
How do you justify the number?
How do you say the price with confidence?

So here's the guide I wish I had when I started.

1️⃣ Stop charging by the hour
Charge for the outcome instead.

Every workflow delivers 3 types of value:
↳ Saves money (cuts costs)
↳ Saves time (frees up hours)
↳ Reduces errors (prevents expensive mistakes)

If you charge hourly, you cap your own income.
The faster you build, the less you make.

That's the wrong game.

2️⃣ The 2 core pricing models
↳ Value-based project pricing (based on business impact)
↳ Monthly retainer (for ongoing maintenance and optimization)

Most freelancers only use one.
The best ones use both.

3️⃣ The 10x ROI rule
This is the number that makes pricing obvious.

Price at 10% of annual savings.

Example:
↳ Workflow saves €12K/year
↳ You charge €1,200

That's a no-brainer for any business owner.

4️⃣ How to price retainers
Simple formula:
↳ Estimate your monthly cost
↳ Apply a 50% margin (double it)

No guessing.
No undercharging.

5️⃣ Present the transformation, not the number
Lead with what changes for them.
Define what "done" looks like before you say the price.
Scope it clearly so there's no room for creep.

Then say the number.

6️⃣ Handle objections by adjusting scope
Never discount your price.

If the budget is tight:
↳ Remove deliverables
↳ Reduce scope
↳ Phase the project

Your price signals your value.
The moment you discount, you signal the opposite.

7️⃣ Turn one-off projects into recurring income
Every project is a door to a retainer.

Offer:
↳ Maintenance
↳ Optimization
↳ Training
↳ Reporting

One client. Multiple invoices.

8️⃣ The V.A.L.U.E. framework
↳ Vision — tie the automation to real business goals
↳ Assess — audit and map the current workflow
↳ Leverage — find the opportunities with outsized returns
↳ Unveil — present the solution with simple ROI math
↳ Elevate — review, optimize, and expand continuously

If you want to learn how to price and sell AI automations properly,
this is where you start.

Full guide + community here:
https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

How To Price Agentic Workflows

Most AI agency owners are doing lead generation.

And wondering why it's getting harder every month.

Here's what nobody tells you:

Lead generation captures demand.
Demand generation creates it.

They're not the same game.

🔸 Demand Generation
↳ Top of funnel
↳ Targets the 95% not ready to buy yet
↳ Builds awareness before people enter buying mode

Tactics: educational content, thought leadership, newsletters, community, ungated videos.

Mindset: give value first, so they remember you when they're ready.

The downside?

Hard to measure.
Slow to impact pipeline.
Requires consistent content.

But the upside compounds:

↳ More inbound over time
↳ Higher quality leads
↳ People already trust you before the first call

🔸 Lead Generation
↳ Mid/bottom of funnel
↳ Targets the 5% already ready to evaluate
↳ Captures interest and moves it fast toward a sale

Tactics: landing pages, demo requests, free trials, direct response ads, downloadable guides.

Mindset: catch who's already interested and move them fast.

The downside?

You can hit your lead numbers and still not grow.
Because you're fishing in a pond you didn't fill.

The real problem most AI agencies have:

They only do lead generation.
They skip demand generation entirely.

So they're competing for the same 5% as everyone else.
On price.
On speed.
On desperation.

The agencies that win long-term do both:

→ Demand gen builds the pipeline over time
→ Lead gen converts what's ready now

Different jobs.
Different timelines.
Both matter.

If you want to learn how to build and scale an AI Agency properly,
check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

Claude Code + Remotion can generate full videos from plain text prompts, and agencies are paying $1-2K for it. If you sell AI services, this is your next offer:

No coding. No editing software. No designers.

Here's exactly how it works:

1️⃣ Install Claude Code desktop app
↳ That's your AI workspace

2️⃣ Give Claude a "Skill" for Remotion
↳ Think of it as an instruction manual
↳ Claude now knows how to build videos

3️⃣ Set a local folder as your workspace
↳ This is where your videos get built

4️⃣ Prompt Claude to create a video
↳ It browses the web automatically
↳ Pulls the content it needs
↳ Builds the full video for you

That's it.

No timeline.
No keyframes.
No back-and-forth with a developer.

Then you iterate in plain English.

Don't like the pacing?
Tell Claude.

Want a different style?
Tell Claude.

That's the process.
Not one-shot. Iteration.

And it goes further:

↳ Add voiceover via ElevenLabs API
↳ Add background music automatically
↳ Chain multiple videos with a single prompt

The output:

→ AI-generated script
→ Voice
→ Music
→ Animations
→ Full end-to-end video

This is the kind of workflow that used to cost $5K and two weeks.

Now it's a skill file and a prompt.

And any agency that still pays a video team to do this manually?
That's your next client.

If you want to learn how to build and sell workflows like this,
check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network
🎥 Full video: https://lnkd.in/dWxU_U7S

I've been on 100+ calls with people who want to start an AI agency. After a while, you see the same story every time: ⬇️

They hate their job. 

Not a little. 
Like, wake up on Monday and feel sick kind of hate.

So they stop trying. 
Why work hard for something you hate? 

And because they stop trying, they stop getting better. 

No new skills. 
No raises. 
Nothing moves.

Then they start looking for a way out. 

Finally the right opportunity shows up... but they don't take it.

Not because it's a bad opportunity. 

But because they've been stuck so long, they stopped believing they can actually make it work.

That's the real problem. 

Not the opportunity. 
The fear of betting on themselves.

The people who made it out didn't have a perfect plan. 

They just said yes when it mattered.

Have you ever regret saying no to an opportunity?

---

PS: Do you like my mom's Pikachu up there?

4 years trying to build a business while working a 9-5.

Nothing worked.

Then the last 12 months changed everything.

And when I look back, it comes down to two things:
↳ Focus
↳ Consistency

But focus is the one that actually shocked me.

Here are the numbers I couldn't ignore:

↳ Knowledge workers check email every 11 minutes
↳ After any interruption, it takes 23 minutes to fully refocus
↳ People switch tasks 1,000+ times per day

That means most people aren't working.

They're restarting. All day. Every day.

Even if you sit at your desk for 8 hours,
your brain never stays on one problem long enough to build anything real.

So here's the system I used to fix it:

1️⃣ Remove inputs
↳ Notifications off
↳ Phone off the desk
↳ No background noise while thinking

Less inputs = more usable brain.

2️⃣ Reduce switching
↳ Pick 3 tasks max per day
↳ Finish one before touching the next
↳ Plan tomorrow the night before

You don't need more hours.
You need fewer resets.

3️⃣ Lower stimulation
↳ Walks without headphones
↳ Meals without content
↳ Driving without podcasts

Silence isn't wasted time.
It's how you rebuild concentration.

4️⃣ Lock in a focus window
↳ Same time
↳ Same place
↳ Same type of work

Your brain learns the pattern.
When you show up, it knows what to do.

This is what allowed me to build something real while still working a 9-5.

Most people can't focus.

Which means if you can...
the gap between you and everyone else grows every single day.

If you want to build an AI Agency around this, connect with others doing the same here: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

5 Google Antigravity use cases no one is talking about.

And they all turn into real agency income:

And if comment "AGENTS", I'll send you the link to the video.

When I first started using Antigravity, I treated it like a cool tool.
Build something fast. Show someone. Move on.

Then I stopped building demos.
And started building offers.

That's when everything changed.

Here's what I packaged into real agency income:

1️⃣ Invoice Processing Automation
↳ Pull invoices, extract data, route approvals automatically
↳ Every business has this problem
↳ Every business will pay to fix it

2️⃣ Lead Enrichment Workflows
↳ Takes a name and a company
↳ Returns full context: role, pain, recent activity
↳ Hours of manual research. Gone.

3️⃣ Facebook Ads Agents
↳ Generates creatives, tests angles, iterates on results
↳ Not one-shot output
↳ A loop that improves itself

4️⃣ SEO Tracker Builds
↳ Monitors rankings, flags drops, surfaces opportunities
↳ Clients don't buy this once
↳ They pay for it every month

5️⃣ After Meeting Sales Assistants
↳ Turns call recordings into follow-up emails, summaries, and next steps
↳ Closes more deals without more calls
↳ The kind of tool sales teams don't give back

The build is only half the work.

The other half is knowing how to package it,
price it, and walk into a client conversation with confidence.

That's what I break down in the full video.

Comment "AGENTS" and I'll send you the link.

And if you want to connect with others building and selling AI services,
check this: https://lnkd.in/d5nE_Fuc

♻️ Repost to help your network

taplio

Giovanni Beggiato's Linkedin Analytics