The Tool Our Media Buying Team Built Themselves
19 okt, 2026 | 5 min read
The Setup No One Talks About
Running localized campaigns at scale is not complicated. But it is slow.
Every location needs its own version, with creative and copy tailored to it. Across a single market, that adds up to a large number of individual pieces that all have to come together correctly before anything goes live.
Michael, Media Buyer at Manna, realized this kind of setup was a regular part of his workflow. Two to three days of careful, repetitive work per campaign. Not because the thinking was hard, but because the execution was manual.
"The logic is the same as any other setup, prepare the creative, ad text, and targeting, then repeat it a few hundred times across every market. Most of the work isn't the strategy, it's making sure everything is built and ready to launch at scale. A day or two just turns into a monkey job of copy pasting details from one place to another, a few hundred times." - Michael Balazh, Senior Media Buyer
The Fix That Never Gets Built
Due to specifics of the setup, existing tools on the market couldn’t solve the problem. The obvious answer was an internal tool. Something that takes the inputs once and handles the rest.
But there is a familiar problem with obvious answers like this. Going through a proper product process, design, build, test, ship, costs more than the manual work it is meant to replace. The numbers do not add up, so the idea never becomes a ticket. It just stays in everyone's head as a "we should really do that someday" conversation.
Most teams have a list of those. Workflows that are too small for a product team to prioritize and too painful to just accept. They sit there, quietly costing time, week after week.
"It's just not a priority, not an everyday task, and it doesn't make sense for a product team to spend resources on this, especially when there are tools on the market that promise to solve this problem. And we were just stuck in a loop where available tools don't fully solve our problem, and making a tool ourselves isn't worth the effort for the output it gives." - Michael Balazh, Senior Media Buyer
How Did Manna Build This Without a Product Team?
Instead of going through the product team, Manna tried something different. The media buying team got access to real hardware capable of running AI tools locally, and Claude Cowork as a building partner.
Cowork sits on your computer and builds working tools through plain conversation. You describe what you need, it helps you build it. No coding background required. What it does require is knowing your problem well, and after months of doing this manually, Michael knew his problem very well.
"Ohhh, I can talk a lot about vibecoding. It took me about half a year to actually be productive with it, so by the time I started this tool, I already knew what I needed. A detailed prompt, explaining what you want, what you have, and how you see it, is key. It's satisfying to see the tool take shape after dozens of iterations, especially past the moments it decides to hallucinate and add more bugs than fixes." — Michael Balazh, Senior Media Buyer
What the Autolauncher Does
The tool is called the Autolauncher.
Michael chooses the locations, drops in the creatives, writes the copy once, and the Autolauncher builds the full campaign structure. Every campaign launches paused by default. A human always reviews it before anything goes live. That was a deliberate decision, keeping a real person in the loop before any budget moves.
The result: what used to take two to three days now takes around ten minutes of prep.
Around 160 launches processed through it so far.
Curious how this could work for your team? Connect with us.
Curious how this could work for your team? Connect with us.
"That was a feeling of relief and satisfaction for creating something that saves me from monotonous, manual work. And it's just nice to see a thing I've spent time building actually working and helping me. " - Michael Balazh, Senior Media Buyer
Why This Matters Beyond The Time Saved
The Autolauncher is useful. But the more interesting part is how it got built.
No backlog. No sprint. No waiting for a team to have capacity. Michael had the problem, he had the tools, and the work shipped.
When the person doing the job every day is also the person building the fix, the output is different. The tool fits the actual work, not a spec written three steps removed from it. Edge cases get handled because the builder has lived them. The feedback loop is immediate.
This is what changes when small teams have access to local AI tools. Problems that were never worth a product team's time become solvable by the people who feel them most.
"It changes how you approach any problem, and opens up room for more automation. If I notice I'm doing the same thing several times a week, I already think about how to automate it. A 15 minute task done twice a week doesn't sound like much, until you count how many of those you actually have in a day." — Michael Balazh, Senior Media Buyer
What Comes Next
The Autolauncher is the first one. It will not be the last.
If your media buying team has a list of workflows that are too small to build and too slow to keep doing manually, that gap is now smaller than it used to be.
The question is who on your team is ready to close it.
Curious what this could look like for your team? Send us a note at sales@wearemanna.io
FAQ
-
What is the Autolauncher?
A locally built internal tool that automates part of Manna's campaign setup. The media buying team inputs the locations, creatives, and copy once, and the tool generates the full campaign structure, always paused for human review before launch.
-
How Long Did it Take to Build?
It was built without going through a product team, using Claude Cowork as an AI building partner. The person who needed the tool built it themselves.
-
Does the Autolauncher make mistakes?
Every campaign launches paused. A human reviews the full setup before anything goes live. The tool handles configuration, the team handles judgment.
-
Can Other Teams Build Something Like This?
Any team with a repeatable workflow that is too small to productize and too slow to keep doing manually is a candidate. The media buying team was first. It will not be the only one.