No. The most interesting data point we have heard on this runs the other way.
Patrick Stuart-Constant is CEO of Sociaaal, which buys mobile apps and scales them with generative AI. On the Mobile User Acquisition Show he described the moment his team got AI production working, and what happened next was not a headcount reduction.
“This is amazing. With two people I can produce 2,000 ads a month. This is huge.”
Two people, two thousand ads a month. That is the promise everyone is currently being sold. Then he tripled the team.
“So we multiplied the number of people in the creative team by three – we went from two to six – and we stayed at 2,000 ads a month for quite a few months. We saw performance, in terms of CTR etc., basically double.”
Read that carefully, because the shape of it is the whole point. Three times the people. The same output. Roughly double the click-through rate.
They did not spend the extra humans on more ads. They spent them on better ones.
Page Contents
Why more AI created more demand for humans
Our host asked the obvious question: was that surprising? Patrick’s answer was simply “Yeah, it was.”
It should not be, once you see what the humans were actually for:
“Keeping humans to inject creativity – finding new marketing angles, new ad concepts, new video types. That’s one of the very important things.”
Generation was never the bottleneck. Judgment was. When you can produce two thousand ads a month with two people, the binding constraint stops being production capacity and becomes the number of genuinely different ideas you can feed the machine. Ideas are the scarce input, and ideas are still made by people.
This is the part that the “AI replaces your creative team” story gets backwards. AI removes the work that scaled linearly with headcount, which was execution. What is left is the work that never scaled with headcount at all, which was taste. You do not need fewer people for that. You need more of a different kind.
Where the winners actually came from
The detail that supports it: the ads that carried the account were not the mass-produced variants.
“The ones that really outperformed – the ones that got to what we call internally ‘unicorn status,’ ads we could run at incredibly low CPIs/CPAs at really high volumes of spend – those were coming from basically brand new ads we were building using AI tools.”
New concepts, not iterations of old ones. And the iterations had a specific weakness:
“It was the multiplying variations. We’ve got quite a systematic process where we test each variable in an ad – dozens of hooks, dozens of actors, etc. for the same winning ad.”
followed immediately by:
“Those videos also have a much shorter lifetime.”
Variants are cheap, fast and fade quickly. New concepts are the ones that scale and last. The first is a machine job. The second is a human job. A team that fires its humans and keeps its generator ends up with an enormous amount of the short-lived kind.
The volume argument, stated honestly
None of this is an argument against volume. Patrick is emphatically for it:
“It has been feasible for many years to make a few dozen good ads. The challenge is producing thousands of high-quality ads.”
And his framing of why volume works is the clearest we have heard:
“If you produce 10,000 awful ads in the totally wrong part of the search space, you won’t perform. If you produce 10 ads in the right part of the search space, you might get a few winners. If you’re producing 100,000 ads in the right part of the search space, at least 10,000 of them are going to be winners.”
Volume multiplies whatever direction you are pointed in. Which is exactly why the humans matter more, not less: they choose the part of the search space you are multiplying inside. Point the machine somewhere generic and it will generate generic at a rate no human team could match.
What a “system” actually means here
Worth quoting in full, because most people mean “a tool” when they say this:
“You need to create the right system. And a system is humans, tooling, workflows, processes… We’ve trained our own data models to allow us to see the signal amongst the noise relatively quickly and cheaply, using a Bayesian approach.”
Humans first in that list. And the scale he is describing required building software just to administer it:
“Originally it was hard to scale efficiently to 5,000 ads, and we had to build internal software to manage, test, and deploy all those creatives on campaigns… Now we’re scaling up towards 10,000 ads. But what really excites me is: how can we scale up to 100,000 ads a month?”
At that volume a new problem appears, which is that you cannot test everything. Their answer is to add a filtering layer before spend:
“So one thing we’re working on as a first filter is using LLMs, based on audience data, to imitate that audience and basically pre-rank ads before we even test them. You produce 100,000 different ads, and the LLM identifies the top 10,000 – those are the ones you actually test on the platform and see how they perform with real humans.”
Note what that is: a curation layer, bolted on because curation is the constraint. Even the proposed fix for having too many ads is more judgment, just automated one level up.
What this means for your team
- Do not plan AI adoption as a headcount reduction. The clearest example we have seen went the other way and got double the CTR for it.
- Move people from execution to concept. The roles that disappear are the ones that scaled with output. The roles that matter are the ones that decide what to make.
- Separate concepts from variants, and resource them differently. Variants are cheap and short-lived. New concepts produce the outliers. Do not let a variant machine crowd out concept work, because variant volume is the easier number to report.
- Expect to build or buy an administration layer. At a few thousand ads a month, managing, testing and deploying creatives becomes its own engineering problem.
- Judge the direction before you multiply it. Volume in the wrong part of the search space is just a faster way to be wrong.
This post is part of our guide to how to choose a mobile UA agency.
Frequently asked questions
Does AI reduce the size of your creative team? Not necessarily. In the clearest documented case we have, a studio producing 2,000 AI ads a month with two people tripled its creative team to six, held output flat, and saw CTR roughly double. AI removes execution work, which raises the value of concept work, and concept work is done by people.
What do the humans do once AI is generating the ads? Finding new marketing angles, new ad concepts and new video types, in Patrick Stuart-Constant’s words. Generation stops being the constraint and the supply of genuinely different ideas becomes the constraint.
Do AI-generated ad variants perform as well as new concepts? In this account, no. The top performers came from brand new ads rather than from multiplying variations on a winner, and the variants had “a much shorter lifetime.”
How many ads a month can a small team produce with AI? Two people were producing 2,000 a month. The same team later scaled toward 10,000 with more people and purpose-built internal software to manage, test and deploy at that volume.
Is producing more ads always better? Only if they are pointed in the right direction. Volume multiplies your direction rather than correcting it, which is why concept quality matters more as production volume rises.
