How a Global Haircare Brand Used AI to Get 227 Creators Live in One Month and Analyze 5,000 Comments

The Challenge

Running 227 nano-creators across two markets creates two problems, and both come from its size. First, sourcing, vetting, and contracting that many creators by hand is a multi-month job. Every creator has to be found, checked for audience quality and brand safety, briefed, and put under contract, and that work multiplies with each name added to the roster. This is the tier that is hardest to scale, because the effort per creator barely drops as the numbers climb.

Second, once they post, thousands of comments come back, far more than any team can read to know what the audience actually thinks. Likes and views tell you a post landed. They cannot tell you whether people trust the product, hesitate at the price, or bristle at being sold to. In a two-market, two-language campaign, that judgement has to be made separately for each audience. What reads as playful in one feed can read as pushy in another, and a single blended engagement number hides exactly the differences a brand needs to see.

Swavy's AI was built for exactly this. Its AI agents got all 227 creators live in a single month, and its sentiment engine read every comment that came back, then turned what it found into direction for the next campaign.

How Swavy's AI Agents Ran It

Swavy is an AI-powered influencer marketing platform. Its AI agents source, vet, contract, and analyse, while the platform gives the team and the client one live dashboard to watch it all happen.

For this campaign, the agents matched creators against the brief's criteria: market, niche, audience quality, and brand-safety signals. They ran compliance and contracting in parallel rather than one creator at a time. Where a human team works through a roster sequentially, the agents worked across the whole list at once, so adding the two-hundredth creator cost no more effort than the second.

Matching is where scale usually breaks down. A nano-creator with the right follower count but the wrong audience adds noise instead of reach, so the AI agents weighed audience composition and authenticity signals for each candidate rather than filtering on follower numbers alone. The dashboard tracked every creator's status, from sourced to approved to briefed to posted, so the team and client could see the roster fill in real time, flag anything off-brief early, and never wonder where a given creator stood.

That produced 150 creators in Saudi Arabia and 77 in the UAE, briefed on a comedic, product-in-hand format and free to interpret it in their own voice, with captions in both Arabic and English. The format was a deliberate choice. It gives nano-creators a frame loose enough to sound like themselves while keeping the product on screen, which is what keeps hundreds of separate posts feeling native instead of templated. Three branded buzz videos anchored the campaign, and the rest ran as organic-feeling UGC across TikTok and Instagram, so the brand showed up in hundreds of individual feeds as something closer to a recommendation than an ad. Work that normally takes months went from brief to live in one.

What Around 5,000 Comments Said

As the posts went live, Swavy's sentiment engine read the comments as they came in, around 5,000 of them across two languages, a volume no human team could work through while a campaign is still running. It did not just tally positive versus negative. It broke down why people felt the way they did, by platform and by comment type. Three findings are worth taking beyond this campaign.

The first sat in the platform gap. The same creative did not land the same way on both. Instagram ran 83.1% positive. On TikTok it fell to 57.3%, with negativity closer to 9%, and the complaint that came up most was that a post felt like a promo. Any brand can use that read. TikTok audiences push back faster on content that looks like an ad, so what works on Instagram often has to be looser and more native to work on TikTok. A single blended score would have buried it.

The second was what actually earned the positive reaction. Looking only at the substantive positive comments, more than half, around 54%, were about the content itself, the humour and the creative. The product, whether love for it or questions about it, made up closer to a quarter. So the comedic format was the engine, not the product pitch. For the next brief, that means protecting the entertainment and the creator's voice rather than crowding it out with product messaging. The examples made the split concrete. A positive comment read, “We have to try this.” A neutral one read, “I'm using the other one, will try.” A negative one read, “Spare us from the ads.”

The third is the one most teams miss. The negativity was small and low-level, split between some mild product criticism and some skepticism about the ads, but it was far more substantive than the praise. 55.7% of the negative comments carried a real, reasoned opinion, against 28% of the positive ones, which were mostly quick reactions. So a low negative score is not a reason to look away. It is where most of the usable feedback actually sits, and reading it is what tells a brand whether to change the product or just the creative.

The Takeaway

The sentiment read became the brief for the next flight. Reception was strongly positive, carried by humour, product love, and affection for the creators, so the goal for the next round was to protect what was working rather than overhaul it. The clearest adjustment was on TikTok, which carried most of the pushback, often some version of “this feels like a promo.” The response was to loosen the brief there, give creators more room, and be less rigid about creative guidelines, so the posts read as native rather than as ads. Instead of guessing at what to change, the team walked into planning with the audience's own reaction already sorted, quantified, and explained.

The broader proof is bigger than speed. Running a campaign at this scale and still hearing what the audience is saying, market by market and platform by platform, is what turns a large campaign from something a brand braces for into something it can actually steer. The same agents that got 227 creators live are the reason no comment went unread, so the brand ended the flight not just with reach, but with a clear, evidence-backed picture of how that reach was received, and a sharper brief for the next one.

Want to see what Swavy's AI agents could do for your brand?

Get in touch with the team today at swavy.app

About the Author

Jacques Saab

Writer

Jacques Saab is a co-founder of Swavy, an AI-powered influencer marketing platform. He leads its agentic AI: the systems that source, vet, and measure creators, turning work that used to take a full team into something a brand can run in minutes.