Deeper Insights, Stronger Impact: Using AI to Scale Expertise
Everyone is talking about AI as an easy button to automate tasks like summarizing documents, handling mundane tasks, or even writing social posts. We’re excited about all of those use cases, but we’re equally excited about areas where AI can make new things possible and expand our capabilities, rather than just make existing work faster. Beyond an “easy button” for data analysis, we firmly believe that AI tools can give us deeper insights into what and how people are talking about the world.
We launched Decibel two years ago to better understand the competitive landscape for public affairs advertising by tracking campaigns across channels and then measuring how effectively they persuaded key audiences. And since we launched Decibel, we’ve used AI tools like machine learning to classify ads by distinguishing public affairs ads from marketing ads or political fundraising from persuasion.
Now, we’re building tools that don’t just classify data based on which words they use — they use AI to understand the context around issues and advertisers to build a much more nuanced map of topics and messaging. We’re already using a new suite of tools that:
- Tag every ad to a more nuanced topic hierarchy. We can distinguish ads focused on carbon neutrality from ads focused on sustainable agriculture, rather than just lumping them into the same “sustainability” category.
- Search for similar campaigns. We can start with one campaign and compare them to similar campaigns based on their context and language, even if they don’t use any of the same words.
- Summarize messages across channels from the entire group of Fortune 500 advertisers.
As a result, we’re equipping our team of analysts to think beyond keywords — to identify the most important ads, not just the ones that use the same words, and connect them to similar campaigns focused on social responsibility and corporate impact. Yes, that means we can move more quickly — but it also gives us a more nuanced view of some of the most complex issues in communications.
Any analyst can quickly differentiate between a marketing ad that sells you a shoe by telling you that it’s made with sustainable fabrics and a reputational ad that defines a retailer’s commitment to improving their supply chain’s impact by using sustainable fabrics. But this kind of distinction has historically really fooled computers because they can’t figure out the different meanings of the same phrase: “sustainable fabrics.”
AI offers us new ways to solve this problem. For example, we’re using a powerful new tool called word embeddings that can understand the similarity between phrases, even if they don’t use the same words. Under the hood, this is part of the technology that powers large language models (LLMs) like GPT to understand the context in our messages to them.
There’s still no substitute for a human in the loop. At their best, these tools can only be as good as the framework that we give them. That’s why these tools only work if experts in media and creative are training the AI. Fortunately, we can lean on the deep and varied experience of the entire BPI team in crafting the best knowledge bases and input data to power our AI approach.
By using AI to scale up the deep public affairs knowledge of our colleagues, we can get a more nuanced view of what advertisers are saying — and what’s breaking through. When we combine that with impact testing to understand what messaging actually persuades people, we know the best way to engage on a novel issue — or avoid pitfalls that other advertisers hit.
As a result, our teams can now quickly discover what’s happening on a given issue to build better, more differentiated creative and targeted media strategies to break through.