Anyone who has built a LinkedIn campaign with dozens, or hundreds, of image and video ads knows exactly how tedious the process is. Every ad set has to be created manually, assets split out one by one between image, video, and carousel formats, and the whole thing repeated across every audience or language variant. It’s slow, repetitive work that eats up hours better spent elsewhere.
Here’s how we replaced that entire manual process using Markifact, and cut a job that used to take hours down to a few minutes of setup and a five-minute wait.
The Tool: Markifact
Markifact is an automation platform built specifically for marketing operations. It connects to the platforms you already use, LinkedIn among them, through dedicated connectors that need to be activated once. Activating the LinkedIn connector is straightforward: you connect, get redirected to LinkedIn’s own sign-in page, authenticate as you normally would, and the connection is live.
From there, you have two ways to work: through an agent inside Markifact itself, or by installing the MCP and driving it directly from Claude. For this walkthrough, we used the agent. Markifact also lets you choose which model and AI tool powers that agent, whether that’s Claude, DeepSeek (notably cheaper in terms of credits consumed), or your own API key plugged in directly.
Writing the Prompt
The single most important part of this workflow is the prompt. The more detailed it is, the fewer questions and headaches you’ll run into later. Our prompt specified the campaign group where the assets should be created, the LinkedIn account name, and instructions to create two ad sets, in this case named BFR and BNL.
For each ad set, we specified the date range, daily budget, target audience, language, and country, along with instructions covering the different asset types and references to the attached images and files to use.
That last part is where the real efficiency comes from: a well-organized Excel file and a consistent image naming convention. Because the whole point of this workflow is that once your prompt, spreadsheet, and image files are properly prepared, the actual campaign creation takes just a couple of clicks.
Running It
With the prompt written, we uploaded the supporting file, an Excel sheet listing ad copy and matching image names, exactly one file, alongside the actual image assets themselves, whose file names matched the spreadsheet precisely. In our example, we used nine images.
One nice touch: when we initially sent the prompt without the file and images attached, the agent caught the omission and flagged it, rather than proceeding with incomplete information. After uploading the correct file and images, mapping BFR to the French-language assets and BNL to the Dutch-language ones, the agent picked up the mistake in our language mapping and asked for confirmation before proceeding.
Since our approval settings were configured to require sign-off by default, the agent paused and asked us to approve both ad sets before creating anything live. We approved, and roughly five minutes later, the job was done.
Verifying the Result
Back inside the LinkedIn platform, we checked the default campaign group and filtered on our test setup. Both ad sets were there, exactly as specified, and every individual ad within them had been created correctly, matched to the right assets and language.
For anyone who has built LinkedIn ads manually, the difference is night and day. What normally takes hours of repetitive clicking, asset by asset and ad set by ad set, was reduced to writing one detailed prompt, uploading a spreadsheet and a folder of images, and approving the result.
If you’re interested in automation for PPC, check out our trainings at clicktrust.be.
