A restaurant owner in Wynwood showed me his invoices before he showed me his menu. The year before, he’d paid an agency for a marketing package. Real money.
The photos were pretty. The captions were polished. Nothing on the register moved.
So he quit posting. His feed sat dead for months while the taqueria two blocks over posted daily and stayed packed.
He didn’t need a lecture about marketing. He needed proof it could work without another bill.
Here’s the proof. We rebuilt his visual brand with free tools and the phone in his pocket. Ninety minutes a week, every week.
Profit went up 25%. Two boring things drove it: the dishes we reshot got ordered more, and promos that used to take two weeks shipped the same day.
The same-day part matters more than it sounds. When a Saturday storm emptied his patio, he had a soup special posted before noon. The old way, that promo dies in an email thread with a designer who answers on Tuesday.
This is the playbook. The exact stack, the weekly routine, where it breaks, and the numbers that prove it worked.
The problem was never the tools
I ran big-data marketing science at Fashion Nova, a brand built on relentless daily posting. The lesson from that scale holds at every scale: consistency beats brilliance.
One great photo does nothing. Decent photos, every day, for months, change what people order.
And consistency isn’t just for the algorithm. Regulars check your feed like a storefront window. A dead feed reads as a dead kitchen.
Owners already know this. What they can’t afford is the production line behind it: a photographer for menu shots, a designer for every promo, a retainer to keep the feed alive. Each step costs money, so each step becomes a bottleneck, and the feed goes quiet right when the business needs it loud.
Generative tools collapse that production line into one person with a phone. The owner becomes the photographer, the designer, and the agency. The skill that used to cost a salary now costs a written prompt.
I covered the budget side of this shift in my piece on moving from ad spend to AI spend. The short version: don’t pay to promote weak images. Make strong images for free first, then decide if you still need the ads.
The $0 stack, tool by tool
Four pieces, each with one plain job. Every piece has a free tier that covers a single-location business.
- Your phone camera. The raw input. Shoot the real dish in daylight, near a window. AI should improve your photos, not invent your food.
- A free image generator. Turns a raw phone shot into a menu-grade image: cleaner lighting, a better table, a seasonal backdrop. Google’s Gemini does this on its free tier, and it follows plain instructions like “make this look like a cookbook photo.”
- A free cleanup editor. Removes backgrounds, fixes color, and resizes one image for every platform. Adobe Express handles all of that on its free tier. Adobe named me an AI Change Maker, and I’d still pick it at $0.
- A free scheduler. Posts while you run the business. Meta Business Suite schedules a full week across Facebook and Instagram at no cost.
That’s the whole stack. No subscriptions, no contracts, no dashboard you’ll forget exists by March.
One more habit makes it work: when a prompt gives you a look you love, save the exact words in your notes app. Reuse them every week. At this size, that saved prompt is most of your brand consistency.
The 90-minute batch day
Tools without a routine become another folder of dead apps. So we built one routine the owner could keep: one block, once a week, ninety minutes, same day every time.
- Minutes 0 to 15: shoot. Ten quick photos of the week’s featured dishes by the window. Phone only. Ugly is fine, it’s raw material.
- Minutes 15 to 50: generate and clean. Run each shot through the generator, fix backgrounds in the editor, save everything to one folder.
- Minutes 50 to 70: write. Captions and one promo for the week. Plain words. Name the dish, name the day, give people a reason to come in.
- Minutes 70 to 90: schedule. Load the week into the scheduler and close the laptop. Done until next batch day.
Batching is the trick. Decisions are the expensive part of content, so we make them once a week instead of every morning.
Ninety minutes buys seven days of presence. And because the pipeline is fast, reacting costs nothing: rain, a holiday, a dead Tuesday. Make the promo, post the promo, same day.
I teach this exact workflow at GP Tuesdays, the free weekly AI training I run for Miami entrepreneurs. It’s our most requested session, which says a lot about why I call Miami an applied AI city. Owners here don’t want theory. They want the thing that sells more food this week.
Where free breaks
I’ll be straight about the limits, because the limits are real. Hitting them unprepared is how owners get burned twice.
- Brand consistency at volume. Free tiers drift. Ten images look great. Two hundred start to look like five different restaurants made them. At that volume, you pay for tools that lock your colors and style, or you build templates and reuse them hard.
- Rights questions. Free tools have terms. Read what they allow for business use before you build on them. And never generate a dish you don’t serve. The photo earns the first order. The real food earns every order after that.
- The work a designer earns. Your logo, your packaging, your core brand system. That work is judgment, not production. Hire a human for it. The money this stack saves is how you pay them properly.
Free is a starting line, not a religion. Graduate on purpose, when a number tells you to, not because a sales page scared you.
The goal stays simple either way: production cost should never again decide whether your business shows up.
Measure orders, not likes
The agency that burned this owner sent reports full of reach and impressions. Pretty charts, zero dollars. We replaced them with two numbers a busy owner can track on paper.
Orders per dish. Before you reshoot a dish, write down how many times it sold last week. Reshoot it, post it, wait two weeks, count again.
That difference is the whole report. Our reshot dishes sold more, and the kitchen noticed before any spreadsheet did.
Promo redemptions. Every promo needs a hook you can count: a code word at the counter, or “show this post.” If nobody redeems it, the promo failed, no matter how many likes it collected.
Likes are applause. Redemptions are revenue.
Tape both numbers to the office wall. If they move, keep going. If they stall, change the dish or the offer, not the font.
That discipline is where the 25% came from. Not one viral post. Dozens of small wins we could count, stacked over a season.
I’m the Fractional Chief of AI for businesses that know they’re behind. The full model is in my essay on the fractional Chief of AI: embed, ship, hand off. I take you from watching AI happen to running on it in 90 days. Bring your menu. Book a strategy call.