Back to blog

AI for social media: gimmick or game-changer? A Belgian bakery tells all

19 May 2026

Wij volgden bakkerij De Wolf 90 dagen lang met een AI-tool. De cijfers: 4u/week bespaard, 38% more reach, 22 new regular customers. But what went wrong? Wait - let me provide a proper translation: We followed bakery De Wolf for 90 days with an AI tool. The numbers: 4hrs/week saved, 38% more reach, 22 new regular customers. But what went wrong?

"AI for social media is hype. I'll only believe it when I see hard numbers." Meet Tom De Wolf, third-generation baker from Sint-Niklaas. We put him under the microscope for 90 days. These are the real results — both what worked and what didn't.

The starting point

Bakkerij De Wolf has been around since 1962. Tom (38) took over in 2018. Three employees, two shops, a loyal customer base. The problem: his social media had been dormant for months.

Before the test:

  • Facebook: 1,840 followers, average of 1 post every 3 weeks, average reach of 230 people per post.
  • Instagram: 720 followers, average of 1 post per month, average reach of 95 people.
  • Time Tom spent on it: "Inconsistent. Good month: 2 hours, bad month: 0."

Goal of the test: could we use an AI tool to consistently publish 4 posts per week on both channels, in his voice, without Tom spending more than 30 minutes per week on it?

Week 1-2: The setup

We did a 45-minute onboarding session. Tom told us:

  • The story of the business (3 generations, artisan, no industrial additives)
  • The target audience (families from the Waasland region, aged 35+)
  • The tone (warm, slightly dry humour, nothing "marketing-like")
  • The product range (breads, viennoiserie, pastries, seasonal specialities)

The AI tool created a week's worth of proof-of-concept content. Tom approved 6 out of 8 posts straight away. Two were revised — one had too formal a tone and another referenced incorrect opening hours. That was the entire setup.

Week 3-13: The results

From week 3 onwards, the system ran fully automatically: 4 posts per week per channel, daily approval by Tom, automatic publishing.

Reach and engagement

  • Facebook reach per post: 230 → 318 (+38%)
  • Instagram reach per post: 95 → 174 (+83%)
  • Number of posts in 90 days: 8 → 96 (12× more)
  • Comments per post: 2.1 → 4.7
  • Facebook followers: +127 (from 1,840 to 1,967)
  • Instagram followers: +204 (from 720 to 924)

Concrete business impact

During the test, Tom kept track of how many customers explicitly said they had found him through social media. 22 new regular customers in 90 days. With an average customer lifespan of 4 years and average spending of €15/week, that represents a rough value of €69,000 over their customer lifetime.

Time Tom spent on it

  • Setup (one-off): 45 minutes
  • Weekly approval: average 22 minutes
  • Taking photos (once a month): 30 minutes

Total over 13 weeks: ~6 hours. Compared to creating the posts himself (~3.5 hours per week = 45 hours over the same period), that's a time saving of ~39 hours.

What went wrong?

To be honest: not everything worked straight away.

Problem 1: Texts too long at the start

In weeks 3-4, the AI was still producing texts of 250+ words. That works for Facebook, but gets cut off on Instagram. After a feedback round, the system learned that Instagram posts should be a maximum of 100 words.

Problem 2: Wrong supplier name

Tom works with a specific mill, and the AI once mentioned a different one (a competitor). Tom caught it during the approval process. Since we added the supplier information to the Brand Brain, it hasn't happened again.

Problem 3: Seasonal mismatch in week 6

In May, the AI suggested a post about "warming winter breads". Tom rejected it. We refined the holiday calendar and seasonal settings — since then, every post has been in line with the time of year.

What Tom would do differently

Tom's own lessons after 90 days:

  1. "Invest 1 hour in the setup." The more precisely you teach the tool your brand, the fewer rejections you'll have later.
  2. "Photos remain my job." AI-generated images are fine for standalone posts, but for specific products he prefers to shoot a batch himself each month.
  3. "Stay in the inbox." AI schedules the posts, but responding to comments and DMs remains the human touch that customers truly respond to.

Conclusion: gimmick or game-changer?

Tom's verdict, in his own words: "Game-changer. At first I thought: another gadget. But the numbers are there. And above all — I'm no longer embarrassed when someone says 'I follow you on Insta'."

Not every business will see the same results. A retail shop or service provider behaves differently on social media than a bakery. But the pattern is consistent: in 90 days you go from "barely present" to "professionally active", without it costing you more than half an hour per week.

Still unsure whether to choose AI, an agency, or do it yourself? Read our comparison and the full cost breakdown.

Ready to let your socials run themselves?

Free forever. Ready in 10 minutes. No credit card.

Start for free