How Osteria del Borgo Cut Platform Commissions by €86,000 Using an AI Marketing Agent

Establishment: Osteria del Borgo Location: Milan, Italy Covers: ~8,500 monthly Problem: Paying €12,000+ monthly in TheFork commissions Solution: Autonomous AI marketing agent Result: €86,000 saved in year one, 63% platform dependency reduction, zero manual marketing work

This is the complete story of how owner Marco Rossi transformed his restaurant from platform-dependent to platform-independent—using an AI agent that works 24/7 while he focuses on hospitality.

The Problem: Trapped in the Platform Commission Cycle

March 2024. Marco Rossi ran the numbers on his Milan osteria and didn’t like what he found.

Monthly Revenue Breakdown:

• Total covers: 8,500
• TheFork bookings: 5,950 (70%)
• Direct bookings: 2,550 (30%)
• TheFork commissions paid: €12,400/month
• Annual commission expense: €148,800

Seven out of ten diners found Osteria del Borgo through TheFork. Those diners cost €2.08 per cover in commissions (Marco’s contracted rate). The math was brutal:

Annual Financial Reality:

• €148,800 to TheFork
• €35,000 to other platforms (OpenTable, The Fork Premium features)
• Total platform tax: €183,800 annually

For a 65-seat restaurant with €850,000 annual revenue, that’s 21.6% of total revenue going to booking platforms.

Marco tried the usual strategies:

• “Book direct” messages on table tents (ignored)
• Website booking form (rarely used)
• Manual email campaigns to past diners (time-consuming, inconsistent results)
• Instagram posts encouraging direct reservations (minimal impact)

Nothing worked. TheFork controlled the customer relationship. When guests wanted to book their next meal, they returned to TheFork—where competing restaurants could outbid Marco for their attention.

The fundamental problem: Marco didn’t own his customer relationships. TheFork did.

The Decision: Deploy an Autonomous AI Marketing Agent

April 2024. After reading about autonomous AI marketing agents, Marco was skeptical but desperate.

His concerns:

• “Another tool requiring daily management?”
• “How can AI truly work independently?”
• “What if it sends inappropriate campaigns?”
• “Will I lose the personal touch our guests expect?”

What convinced him: A single promise.

“Your AI agent will import guest data, launch campaigns, and optimize strategies autonomously. You won’t touch it. It runs 24/7 independently.”

Marco decided to test this claim.

Implementation: 72 Hours from Setup to Autonomous Operation

Day 1: Integration

• Connected AI agent to TheFork API (guest data import)
• Integrated with reservation system and POS
• Synced existing email list (380 addresses)

Day 2: Initial Sync The AI agent autonomously:

• Imported 4,200 guest profiles from TheFork
• Downloaded complete reservation histories
• Analyzed spending patterns and visit frequency
• Created initial behavioral segments:
  • VIP diners (€80+ average check, 4+ visits)
  • Regular visitors (3+ visits in 6 months)
  • One-time diners (single visit, 60+ days ago)
  • Lapsed guests (visited once, 90+ days inactive)

Day 3: First Campaigns Launch Without Marco’s input, the AI agent:

• Sent win-back emails to 840 lapsed guests
• Launched birthday campaign for 65 guests with upcoming birthdays
• Created VIP offer for top 120 spenders

Marco didn’t write these campaigns. Didn’t schedule sends. Didn’t segment customers. The AI did everything autonomously.

His reaction: “I honestly didn’t believe it would work without constant tweaking. I was wrong.”

First 30 Days: Autonomous AI Proves Itself

The AI agent worked 24/7 without human intervention. Here’s what it accomplished independently:

Campaigns Launched Automatically:

• 15 win-back campaigns (lapsed guests)
• 8 birthday/anniversary campaigns
• 4 VIP exclusive offers
• 12 “direct booking incentive” campaigns
• 3 seasonal menu preview campaigns

Total emails sent: 6,240 Human time required: 0 hours Marco’s involvement: Checking dashboard twice weekly

Results - Month 1:

• Direct bookings: 2,550 → 3,180 (+630 covers)
• TheFork bookings: 5,950 → 5,420 (-530 covers)
• Platform dependency: 70% → 63%
• Commissions saved: €1,100

Not revolutionary. But completely autonomous.

Marco’s takeaway: “I spent zero hours on marketing. The AI handled everything. That alone was worth it.”

Months 2-6: AI Optimizes Independently

This is where autonomous AI differs from marketing automation.

Marketing automation runs the campaigns you configure. AI marketing agents optimize what works and kill what doesn’t—independently.

What Marco’s AI Agent Did Autonomously:

May 2024:

• Identified that win-back emails with €15 prepaid bonus converted at 12% (vs. 6% for no incentive)
• Automatically shifted strategy: All win-back campaigns now include prepaid offer
• Result: Win-back conversion doubled

June 2024:

• Discovered VIP guests responded better to “exclusive tasting menu preview” than discounts
• Autonomously adjusted VIP campaigns to emphasize exclusivity over savings
• Result: VIP engagement up 34%

July 2024:

• Detected that Tuesday/Wednesday lunch was weakest period
• Launched autonomous proximity campaigns (SMS to nearby guests)
• Result: Weekday lunch covers up 18%

August 2024:

• Analyzed that guests who visited 3+ times had 85% retention rate
• Created automatic “third visit reward” campaign triggering after second reservation
• Result: Third-visit conversion rate increased from 45% to 68%

Marco’s involvement in all these optimizations: Zero hours.

The AI identified problems, tested solutions, measured results, and scaled what worked—all autonomously while Marco ran his restaurant.

Month 6 Results: The Numbers That Matter

September 2024 Performance:

Booking Sources:

• Direct bookings: 4,680 (55%)
• TheFork bookings: 3,820 (45%)
• Platform dependency: Down from 70% to 45%

Financial Impact:

• TheFork commissions: €12,400/month → €7,950/month
• Monthly savings: €4,450
• Six-month savings: €26,700

Guest Database Growth:

• March 2024: 380 email addresses
• September 2024: 8,200 owned guest profiles
• AI-driven database growth: 2,058% increase

Campaign Performance (All Autonomous):

• Total campaigns launched: 127
• Total emails sent: 41,500
• Average open rate: 42%
• Conversion to reservation: 11.3%
• Direct marketing-driven covers: 2,140 (in 6 months)

Staff Time Spent on Marketing:

• Before AI: 15-20 hours weekly
• After AI: 0 hours weekly
• Time freed: 360-480 hours in 6 months

Year One: €86,000 Saved, 63% Platform Dependency Reduction

March 2025 - One Year Results:

Platform Dependency:

• Starting: 70% TheFork bookings
• After 12 months: 26% TheFork bookings
• Reduction: 63% decrease in platform dependency

Financial Performance:

• Year one commission payments to TheFork: €62,400
• Previous year commission payments: €148,800
• Savings: €86,400

Guest Database:

• Owned profiles: 12,400 (up from 380)
• Active guests (booked in last 90 days): 6,800
• VIP segment (€80+ average check): 940

AI Agent Performance:

• Campaigns launched: 540+ (autonomous)
• Emails sent: 187,000+
• Direct bookings driven: 11,200+
• Revenue attributed to AI campaigns: €420,000+

Marco’s Time Investment: Checking dashboard 2x weekly (30 minutes total per week)

The Transformation: From Platform-Dependent to Platform-Independent

Before AI Agent (March 2024):

• 70% of bookings through TheFork
• €148,800 annual commissions
• 380 owned guest profiles
• 15-20 hours weekly on manual marketing
• No systematic guest relationship management
• Complete dependence on platform for customer acquisition

After AI Agent (March 2025):

• 26% of bookings through TheFork (strategic, not dependent)
• €62,400 annual commissions (58% reduction)
• 12,400 owned guest profiles (3,163% increase)
• 0 hours weekly on marketing
• Autonomous guest relationship management
• Independent customer acquisition capability

The difference: An AI marketing agent working 24/7 autonomously.

What Made This Work: True Autonomous Operation

Marco’s success wasn’t about better campaigns or smarter strategy. It was about AI that actually works independently.

His AI agent:

1. Imports data autonomously - Pulls guest information from TheFork daily, no manual CSV uploads
2. Segments intelligently - Analyzes behavior patterns and creates segments automatically
3. Launches campaigns independently - Decides when to send win-back emails, birthday offers, VIP perks
4. Tests continuously - Tries different subject lines, send times, offer values
5. Optimizes automatically - Doubles down on what works, kills what doesn’t
6. Tracks real ROI - Measures actual reservations generated, not vanity metrics
7. Runs 24/7 - Never stops, never forgets, never needs management

Marco never:

• Wrote campaign copy
• Scheduled sends
• Segmented customers manually
• Analyzed performance reports
• Adjusted strategies

The AI did everything. Autonomously.

Marco’s Advice for Other Restaurant Owners

After one year of autonomous AI marketing, Marco has strong opinions:

“Stop thinking of this as a tool. It’s an employee that never sleeps.”

“I was skeptical about ‘set it and forget it’ claims because every tool I tried still required constant work. But autonomous AI is different. It genuinely runs independently. I check results twice weekly, but the AI doesn’t need my input to function.”

“The savings are real, but the time savings matter more.”

“€86,000 saved is fantastic. But 20 hours per week freed up? That’s transformational. I’m in the dining room during service, not in the office managing campaigns. That’s worth more than the commission savings.”

“Platform dependency reduction happens systematically, not overnight.”

“We went from 70% to 26% platform bookings in 12 months. That’s slow enough to be sustainable, fast enough to matter financially. The AI handles it autonomously—continuously importing guests from TheFork, converting them to direct bookers through targeted campaigns.”

“If you’re paying €5,000+ monthly in platform commissions, you can’t afford to ignore this.”

“Do the math. If an AI agent costs €500-1,000/month and saves you €5,000+ in commissions, that’s 5-10x ROI. Plus you get 15-20 hours weekly back. The decision isn’t whether to deploy AI—it’s why you haven’t already.”

The Bottom Line

Osteria del Borgo’s transformation wasn’t magic. It was autonomous AI marketing.

An agent that:

• Imported 12,400 guest profiles from platforms
• Launched 540+ campaigns independently
• Drove 11,200+ direct bookings
• Saved €86,400 in commissions
• Required zero manual management

All while Marco focused on running his restaurant.

That’s the promise of autonomous AI marketing. Not better tools. Not smarter automation. AI that genuinely works independently while you focus on hospitality.

Ready to cut your platform commissions?

Discover how your autonomous AI marketing agent can reduce platform dependency systematically—explore Caramel’s AI marketing agents or learn about the Signature Concierge Service for Michelin-starred restaurants.

---

Real restaurant. Real numbers. Real autonomous AI. Osteria del Borgo reduced TheFork dependency by 63% and saved €86,400 in year one—while spending zero hours on marketing.