“Before Civis, it took months and a whole team to figure out what was moving the needle. Now, we can see what’s working almost instantly.” - Kate McKenney, Managing Director, Climate Advocacy Lab
Summary of Outcomes
- 13,000 ads categorized using AI in a matter of hours
- 70% accuracy on first pass, improving to 88% with light tuning
- Benchmarks on CPC, CPM, and CTR unlocked across campaigns
- Live data feeds for real-time insight and community collaboration
The Challenge
The Climate Advocacy Lab exists to equip the U.S. climate movement with evidence-based insights, tools, and data. But with over $2 million in climate-related ad spend to analyze and 13,000+ individual ads from multiple partners and platforms, the process of categorizing and comparing campaign performance was unscalable by hand.
Temporary human coders found the work time-consuming and error-prone, and the Lab needed a more efficient way to process and structure anonymized data. Without a consistent taxonomy or aligned campaign frameworks across organizations, community learning was stalling under the weight of inconsistent inputs and fragmented workflows.
The Solution: Civis AI + Claude Sonnet = Faster Insight at Scale
As members of the Movement Cooperative, the Lab already had access to the Civis Platform, a secure, centralized data environment. Civis integrated Anthropic’s Claude Sonnet model to help categorize thousands of ad records using natural language prompts and structured tagging rules. Civis leverages Amazon Bedrock to provide access to Claude, keeping the Lab's data private, along with Amazon Redshift for secure, scalable storage.
After a single round of prompt engineering, Claude was able to match the accuracy of human coders while processing the dataset in just hours, enabling real-time ingestion, benchmarking, and synthesis of digital ad performance.
Civis also provided a SQL AI Assistant for users to ask complex data questions in plain English. No advanced coding required. What once took hours could now be done in minutes, and ad performance insights could be shared quickly across the community.
The Results: From Insight Lag to Instant Feedback
Civis AI transformed the Lab’s ad data into a live, queryable asset. Now, participating groups can see how their creative compares against community benchmarks in real time, gaining new insight into which messages resonate most effectively and where ad dollars go furthest.
Before Civis, performance analysis took months of manual effort. Today, the Lab receives live input feeds, deeper regional insights, and real-time reports on how ads are performing. This helps their community refine storytelling, support environmental justice, and scale what works.
As climate narratives evolve and resources tighten, Civis AI ensures advocacy efforts stay adaptive, evidence-based, and impactful.
To see and hear more about Climate Advocacy Lab's application of Civis AI, check out the "Civis Lunch and Learn" video below.
