An agency I work with needed deck graphics. Device mockups, social frames, ad units, the connective tissue that makes a pitch deck feel like the campaign already exists. I’ve made hundreds of these. Every pitch needs them, every new brand resets the board, and the work is really just the same shapes wearing new colors.
For twenty years the way to answer that request was to make the graphics. Open the file, place the brand, export the batch, send the folder. Bill for it, and wait for the next pitch to send the same request wearing a different logo.
This time I didn’t send a folder. I sent a URL.
Behind it is a small studio I built for them: you paste in a brand kit (colors, type, logo), write a few lines of brief, and it composes the full set. Twenty-seven formats, phones and billboards and social frames, each assembled to the brand and exported as clean transparent PNGs their designers drop straight into the deck. The whole team can work in it at once. Nobody needs me in the room.
I keep writing about the moment a task turns out to be a pattern wearing the disguise of a one-off. What was new here was whose pattern it was. This one belonged to the client, and when the pattern belongs to the client, honesty gets expensive. A quiet share of the business of design has always been the business of billing for repetition. The brand changes, the shapes don’t, and the invoice rides the difference. That arrangement made sense when building a tool took a quarter. It stopped making sense the week the tool took a few evenings.
Building it was still design work, aimed at a different altitude. Nobody hands you a spec for a machine like this. The work was deciding: which twenty-seven formats earn a slot, how a brand kit maps onto each one, how far you can push a layout before it stops being on-brand, which controls exist and which temptations deliberately don’t. Every one of those calls is taste, encoded. I wrote in January that the goal of a brand system is to make the easy thing to make also the on-brand thing to make. A machine like this is that sentence, compiled. The craft didn’t leave the work. It moved into the walls.
I’ll admit the fear, because every designer I describe this to arrives at it within a minute: if the client owns the machine, why do they need you? I had the same flinch… it lasted until the machine had been in their hands for a while, and what came back wasn’t silence; it was different requests. Another team wanted a version for their own workflow. The pitch team wanted new formats added, which is a judgment call, which is mine. Someone needed a review system, then a small dashboard, each one a machine-shaped request from people who’d learned what a machine feels like to own. In twenty years, no folder of files ever generated a follow-up on its own. Files close a project… a machine opens the next one.
There’s a limit here, and it’s worth being precise about. The machine makes the middle of the work: the twenty-seventh social frame, the mockups that carry a story, the volume nobody should be handcrafting. The edges stay handmade. The idea the deck is selling, the flagship image, the line everyone remembers: none of that comes out of a template, and a machine full of beautiful middles can’t rescue an empty edge. If anything, automating the middle raises the price of the edges, because the edges are what’s left to be good at.
What changes most is what the client is buying. They used to buy outputs, priced by the batch, and every batch made me a vendor with a deadline. Now they own the capacity, and what they buy is the judgment around it: what the machine should make, what it should refuse, when it needs to grow a new format, why an output that looks fine is quietly “off.” The author has to stay attached, because encoded taste rots when nobody updates it. That’s a different invoice, and a better relationship.
The requests from that client have changed shape since. Less “can you make these” and more “can you make the thing that makes these.” I don’t think they decided to change; the machine taught them what was possible, and the old asks started to feel like asking a well for a single glass of water. For twenty years my projects ended with a folder changing hands. Lately they end with a URL that makes more. I know which ending gets me invited back.