The Franchise Screen Problem and the End of the Weekly Zip File

A regional marketing coordinator used to spend Thursday afternoons the same way: open the shared drive, download a zip file of menu board PDFs from the agency, rename them for each location, and email store managers with instructions on which file to push to which screen. If a price changed on Friday, the whole loop started over. A new generation of tools is dismantling that workflow, generating on-brand screen layouts from a short written brief and then scheduling and distributing them automatically across every location.

This matters because the people running screens at twenty, two hundred, or two thousand locations rarely have enough designers on staff, and the agency retainer that filled the hole has rarely scaled well. A small team now has to decide how much of the design work to hand to AI, where human judgment still earns its keep, and what to stop paying for entirely.

Decide Whether to Keep the Agency on Retainer

The retainer that produced the weekly PDF drop was paying for two things: a designer's hands and a project manager's inbox. The hands made the layouts. The inbox fielded one-off requests from the regional VP who wanted a soft-launch promo in three markets by Tuesday. Generative tools can replicate the first job in minutes.

They can't replicate the second as reliably, because judgment calls about brand, timing, and local context still land on a human. One of the newer entrants is an AI-powered signage platform that just launched for multi-screen businesses, and it pitches exactly this brief-to-screen idea: describe what the display should say, let the system design and schedule it, play it on whatever hardware is already on the wall.

The practical move is to split the retainer rather than cancel it. Keep a designer on call for the quarterly brand refresh, the new store opening, and campaigns that need an actual concept. Hand the recurring layout work, which is most of it, to the AI tool. Industry roundups of creative automation describe exactly this split: master designs become templates, and AI generates the hundreds of localized variants that used to eat a junior designer's week.

Decide What AI Actually Writes to the Screen

Letting a model generate a layout is not the same as letting it decide the content. The copy on a menu board is partly marketing and partly a regulated disclosure, and the two have different tolerances for improvisation.

The sensible division looks like this:

• Layout and styling. Let the AI handle composition, hierarchy, color weighting, and format variants (vertical, horizontal, with and without a QR code). The time savings live here.
• Price and item names. Pull these from the system of record, not from a prompt. A model that paraphrases "spicy chicken sandwich" into "spicy chicken melt" is a recall problem waiting to happen.
• Required disclosures. Calorie counts, allergen notes, and other mandated text belong in a locked field the model can't rewrite. Multi-location operators already know that menu labeling rules extend to menu boards and drive-thru boards, and those requirements vary by jurisdiction. Your automation has to respect that or it becomes a liability.

Decide Which Screens Get Which Variant

The old zip-file workflow treated every location as the same location. One PDF, every store, same hours. That rarely matched reality, but it was the only thing a small team could execute. AI-generated variants remove the constraint, which forces a question nobody had to answer before: how granular should the targeting actually get?

Daypart scheduling is the easy win, because breakfast, lunch, and late-night menus already differ. The regional calls are harder. Think of a promotion that lands in college markets but flops in family ones, or a price that runs a dollar higher in one metro than another, or a limited-time item you can only pour where the supply chain can keep up.

Each layer of variation is cheap to produce now and expensive to govern. Pick the two or three dimensions that move revenue and ignore the rest, at least until you have the staffing to audit what's playing where.

Decide Who Approves the Output Before It Goes Live

The first draft from a generative tool is usually close enough to feel finished and wrong enough to embarrass you if nobody looks. A phantom apostrophe. A promotional price that contradicts the current POS. A background image that reads fine on a monitor and turns to mush on a 55-inch screen in afternoon sun.

Small teams tend to pick one of two approval models. The first is a single reviewer who signs off on everything before it publishes, which is safe but slow and reintroduces the bottleneck AI was supposed to remove. The second is tiered approval: templates get reviewed once at the brand level, and individual variants publish automatically as long as they stay inside those templates.

Tiered approval scales. It also requires discipline about what counts as a template change versus a content change, and that distinction is worth writing down before the first mistake forces the conversation.

Whether a given team adopts a specific tool or not, the direction is set. The weekly zip file was a workaround for a design bottleneck that no longer exists. What's left is governance, not production: who owns the brand, who owns the data, and who gets to say the screen is ready.