Agency AI: GreenThreads’ 2026 ROAS Strategy

Every agency in digital advertising is looking for an edge. We’re finding it by applying agency AI agents to client media data for PPC optimization which completely changes how we manage and tune campaigns. It gives us a much deeper, more granular read on what’s driving performance and where the real opportunities are. So, what does this actually do for a client’s advertising spend?

Campaign Teardown: Enhancing E-commerce Conversions with AI

Let’s break down a campaign we ran for a direct-to-consumer (DTC) apparel brand called “GreenThreads,” which makes sustainable fashion. In a crowded market, they needed to grow online sales and their customer base. The big goal was hitting a 4x ROAS floor while keeping the Cost Per Lead (CPL) for newsletter sign-ups in check. We had six months, from January to June 2026, and a $300,000 budget to spend across Google Ads and Meta Ads.

Initial Strategy and Creative Approach

We started with a multi-channel strategy targeting environmentally conscious shoppers aged 25-45. On Google Ads, we went after non-branded keywords like “sustainable clothing” and “eco-friendly fashion,” along with “organic apparel.” For Meta Ads, we built lookalike audiences from their existing customer data and also targeted interests like sustainability, ethical consumerism, and outdoor activities. Creatively, we used high-end product shots in natural settings, with messaging that played up GreenThreads’ commitment to ethical sourcing and its environmental impact. This led to three main creative angles: “Style with a Conscience,” “Dress Sustainably,” and “Fashion That Gives Back.”

Targeting and Ad Placement

The Google Ads campaigns were a mix of Search, Shopping, and Display. We used Search for long-tail keywords and Shopping to show off individual products, while Display targeted custom intent and in-market segments. On Meta, we ran ads in Facebook and Instagram Feeds, Stories, and Reels. We put 60% of the budget on Meta at first because fashion is so visual, leaving 40% for Google to capture bottom-of-funnel conversions. We targeted the entire United States but paid closer attention to cities like Portland, Oregon, and Austin, Texas, where people tend to engage more with sustainable brands.

What Worked: Early Wins and Data Signals

The first two months gave us some good signals. The Meta Ads campaigns, especially on Instagram Stories, did really well. Short, dynamic videos of models wearing the GreenThreads line in daily life pulled an average Click-Through Rate (CTR) of 1.8% and a $45 Cost Per Acquisition (CPA) for purchases. Google Shopping was also solid, holding a steady Return on Ad Spend (ROAS) of 3.5x. The creative theme “Style with a Conscience” was the clear winner, getting 20% more engagement on Meta than our other themes. We also noticed that ad copy on Google Search with specific certifications, like “GOTS certified organic cotton,” got better quality scores. This lines up with what we’re seeing everywhere (and what a recent IAB report confirms): consumers want transparent sustainability claims.

What Didn’t Work: Identifying Underperformance

But it wasn’t all perfect. The Google Display Network campaigns were a problem. They got tons of impressions, over 5 million in the first two months, but the CTR was a dismal 0.3% and the CPL for newsletter sign-ups hit $18, way above our $10 target. A lot of those impressions were coming from junky mobile apps with questionable relevance. On Google Search, some of our broad match keywords were driving traffic that bounced immediately (over 70%) without converting. And on Meta, the “Fashion That Gives Back” theme bombed with a 0.9% CTR and a $60 CPA. Turns out the audience cared more about product benefits and personal style. It was also obvious that static image ads on Instagram Stories just couldn’t compete with the swipe-up rates of video ads.

Optimization Steps: The Role of Agency AI Agents

This is where our agency AI agents came in. We spun up a suite of our custom-built AI models, which plug directly into the client’s CRM and advertising platforms. One of these agents is built specifically for media data analysis, and it just watches performance metrics 24/7 across every campaign and ad set, flagging weird patterns and new opportunities as they happen. It was digging much deeper than just automated bidding, getting into user behavior, watching for creative fatigue, and seeing how different audience segments were responding.

1. Granular Placement Optimization: Our AI agent immediately flagged the bad Google Display placements on those mobile apps. It built a list of over 50 low-performing app categories and specific mobile app IDs that were just burning cash, and recommended we exclude them. That one change cut our wasted spend on the Display Network by 15%.
2. Keyword Refinement and Negative Keywords: The AI chewed through the search query reports from Google Search, matching them against conversion data. It found a bunch of broad match terms that were bringing in window shoppers, like “sustainable fashion trends.” Since those clicks weren’t leading to sales, the agent suggested adding negatives like “trends” and “ideas” to sharpen our focus on people who were actually ready to buy.
3. Dynamic Creative Optimization (DCO) and Sentiment Analysis: On the Meta Ads, our AI used natural language processing to read user comments and track engagement. It confirmed what we suspected: “Fashion That Gives Back” had low positive sentiment. So the system automatically shifted more budget to the “Style with a Conscience” and “Dress Sustainably” themes. At the same time, it started generating new variations of our best video ads, testing different call-to-actions and opening hooks in a continuous DCO process that A/B tested headlines and visuals based on what it predicted would perform best.
4. Predictive Budget Allocation: Using a mix of historical conversion data and real-time signals, the AI agent started predicting which campaigns had the highest chance of converting at any given hour. It then adjusted the daily budgets on its own, pulling money from underperformers (like certain Google Display placements) and pushing it to winners (Instagram Stories, Google Shopping). This constant shifting maximized efficiency and made sure the budget was always working as hard as it could.
5. Audience Segmentation and Refinement: The AI also dug into our Meta audiences and found something interesting. A very specific lookalike audience segment, based on people who had bought certain products like organic cotton dresses, had a much higher ROAS (5.5x) compared to our broader interest-based audiences (3.8x). The agent’s recommendation was simple: increase the bid multipliers for this high-value group and start building out other, similarly granular lookalike audiences.

Results After Optimization (Months 3-6)

Putting these AI-driven optimizations into play completely turned the campaign around. The CPL for newsletter sign-ups plummeted and the overall ROAS blew past our initial target. These adjustments, all coming from the continuous analysis from our agency AI, really showed what data-driven decision-making at this scale can do.

The remaining $150,000 of the total campaign budget was spent during months 3-6. In that period, the campaign generated $735,000 in revenue, giving us an impressive 4.9x ROAS. The CPL for newsletter sign-ups dropped 30% from its initial average to $6.25, well below our $10 target. The overall CTR jumped by 35% across both platforms, mostly thanks to the better creative targeting and dynamic adjustments.

I’ve learned that just having AI tools doesn’t get you very far. The real intelligence is in how you configure them and the quality of the data you feed them. Poor data just gives you poor results. Our team spent a ton of time on data integrity and setting up the AI agents with very specific parameters, and I’m convinced that was the difference-maker for this campaign’s success.

Lessons Learned and Future Implications

This campaign taught us a few things. For one, agency AI agents can process and act on huge amounts of media data way faster than any human team which lets us optimize at a level of detail we just couldn’t reach before. We also saw that you have to be testing creative all the time, and AI-driven DCO is the best way to do it. Creative fatigue will kill a campaign, and this is how you fight it. Finally, the initial setup and ongoing monitoring of the AI is everything. The automation does a lot of work, but you need a person watching to make sure the models don’t start drifting or making bad calls based on weird data. For this GreenThreads campaign, we did weekly checks on the AI’s recommendations and actions to make sure they were still lined up with the client’s bigger business goals.

For agencies, the future of PPC optimization is clearly going to involve more advanced AI agents. We’re already experimenting with predictive modeling for customer lifetime value (CLTV) to inform bidding strategies, which is a big step past just optimizing for a one-time conversion. It means we’re optimizing for long-term relationships. This requires a much tighter integration between the AI, CRM systems, and analytics platforms to get a complete picture of customer acquisition and retention.

Using agency AI agents successfully in this campaign shows a clear path to better PPC optimization. When agencies can analyze media data this closely and act on the insights this quickly, they can produce real results and turn ad spend into actual client growth.