AI Travel Scout — demo summary (for the presenter)

The point of this demo

Mateusz's course built this exact system by hand over 4 evenings: ask ChatGPT, copy JSON/prompt/code, paste into n8n, run, it breaks, screenshot back to ChatGPT, paste fix, repeat. Dozens of manual steps, manual VPS deploy.

Here an agent with n8n access got the goal "build Travel Scout from form to email" and did it itself: created both workflows, ran test executions, read the errors, fixed them, re-ran until green. Same result, no human in the copy-paste loop.

What to show, in order

The two workflows in n8n ([adres-twojej-instancji-n8n]:5678) - AI Travel Scout — Intake (5 nodes): form → AI parse → normalize → Data Table insert - AI Travel Scout — Monitor Ranking Gmail (20 nodes): schedule → read table → filter → SerpApi → rank → guide → advisor → Gmail → update table
The Data Table travel_scout_demo, row id 2 — show how the natural-language request became structured fields, then got enriched with top_hotels, best_hotel, destination_guide, recommendation_body, status=sent.
The real email in the inbox (twoj-testowy-adres@example.com), subject "Rekomendacja city break w Barcelonie - czerwiec 2026" — full PL recommendation with Hotel 1898, TOP 3 alternatives, Gaudí facts, day plan.
The 3 bugs the agent hit and fixed by itself: form field mapping, stale node references, Gmail API disabled in Google Cloud. This is the real story — the agent debugged from execution logs, not from a human.

The one-liner

"Four evenings of copy-paste with a chatbot vs one agent run that builds, tests, debugs and delivers a working email. That's the jump from working with a chat to working with an agent."

To run it live

Activate AI Travel Scout — Monitor Ranking Gmail (schedule is every minute), wait ~1 min for the email, then deactivate it. Both workflows are left inactive by default.