Applied AI

AI works when it saves you hours, not when it generates slides.

Petit Roig implements real AI inside your processes — I don't sell conceptual workshops or presentations. Audit, custom agents, automations with n8n, and training so your team owns the system.

90 days
Measurable ROI from the first implementation.
2-3h day
Typical saving per person in a small team.
UCM
Master's in AI for Communication and Media.
My thesis

Three things AI actually solves for you.

And many others it doesn't. The important thing is knowing the difference.

01

Repetitive tasks with structured data.

Classifying emails, enriching leads, generating draft offers, transcribing calls, extracting data from invoices. Work you already know how to do, but that's stealing 2–3 hours daily from your team as a whole. This is where AI is a massive lever.

02

First-level support and triage.

Agents that answer 70% of visitor questions, filter leads before they reach a human, and route real cases to whoever handles them. It's not about "replacing", it's about not losing the client because nobody answers on a Sunday at 10pm.

03

Insights you don't have time to look at.

Automatic campaign summaries, anomaly detection, weekly recommendations based on your own data. What a senior consultant would tell you if they'd spent a day reviewing your dashboard. Documented and delivered, not produced by an anonymous chat.

Real use cases

Concrete examples of what I implement.

These aren't hypothetical — they're workflows I've deployed, adapted to your context. The examples below are indicative of time saved.

Support agent

Chatbot with real context from your company

Answers frequent questions by reading your website, catalogue and internal documentation. Learns from every conversation. Routes to a human when it detects purchase intent or a complaint. Integrated with your CRM.

↓ ~15h/week openai · n8n · hubspot
Outbound email

Mass personalisation without looking like spam

I enrich leads with public info (web, LinkedIn), generate a personalised first line for each one, and integrate with your email tool. We get 3× the opens of a generic sequence.

↑ 3× opens anthropic · clay · lemlist
Reporting

Weekly reports written automatically

Every Monday you'll get a PDF in your inbox with the week's metrics, detected anomalies, and two or three concrete recommendations. Based on your real data from GA4, Google Ads and Meta.

↓ ~4h/week gpt · ga4 · notion
Content

Idea pipeline for social & blog

Every Monday, 10 proposed headlines adapted to your tone, sector and SEO objectives. Not to publish blindly — so you have something to choose from. You approve, the team executes.

↓ ~6h/week claude · seranking · notion
Operations

Invoice and delivery note processing

Reads PDFs, extracts structured data, validates against orders, syncs with your ERP. Alerts on discrepancies. Especially useful for distribution, logistics and paper-heavy services.

↓ ~20h/week gpt-vision · n8n · holded
Voice & transcription

Transcription and analysis of sales calls

Every sales call is transcribed, objections are analysed, insights are extracted and saved to the CRM with tags. Spots patterns — what helps close, what doesn't.

↑ More conversions whisper · claude · hubspot
How I implement it

I connect your stack with n8n and models.

I don't sell you magical SaaS. I build visible, documented workflows that are yours — that you can maintain or migrate without depending on me.

Trigger
Email · form · schedule
Enrich
CRM · web · LinkedIn
AI Model
GPT · Claude · Gemini
Transform
Validate · filter · tag
Destination
CRM · Slack · email
workflow.n8n · example
// Simplified example: incoming lead triage
{
  trigger: "webhook · web form",
  enrich: ["clearbit", "hubspot.find_company"],
  classify: {
    model: "claude-sonnet",
    prompt: "ICP match? intent level? recommend next step"
  },
  route: {
    hot: "slack.sales + calendly auto-invite",
    warm: "email sequence · 5 steps",
    cold: "newsletter · tag"
  }
}
Stack

Tools I actually use.

No religion about tools — I use the best one for each case. These are the ones I use most.

🤖
OpenAI
GPT-4o, Whisper, Vision
🧠
Anthropic
Claude Sonnet, Opus
⚙️
n8n
Visual orchestration
📊
HubSpot
CRM and attribution
📝
Notion AI
Docs and knowledge base
🔗
Zapier
Simple integrations
📞
Twilio
Voice and SMS
🔎
Perplexity
Automatic research
📈
GA4 + GTM
Full analytics
💬
Slack API
Internal alerts
🐙
Supabase
Data & vectors
🏷️
Clay
Lead enrichment
What sets me apart

Typical AI consultant vs. Petit Roig.

The market is full of AI promises. Here's what you usually find — and what you'll find with me.

Typical AI consultant
  • Sells workshops and presentations; zero execution
  • Recommends 20 tools without knowing if they fit
  • Doesn't understand your business reality
  • Disappears once the "strategy" is delivered
  • Doesn't measure the time actually saved
  • Leaves you depending on them
Petit Roig
  • 15 years executing real growth, not theorising
  • I pick the minimum viable tool, not the shiniest one
  • Real audit of your operations before proposing anything
  • Implementation + documentation + training for your team
  • We measure hours saved at 30, 60 and 90 days
  • The goal is your autonomy, not dependency
Next step

Start with an audit. 90 minutes, concrete.

I'll show you three concrete AI levers for your current operations, with an estimate of hours saved. If we decide to implement, we deduct the audit from the project.

Book the AI audit See all AI services