The AI hype is hard to miss.
From AI-generated ad creative to AI-optimized ad targeting, this new frontier of technology promises to make ads better and marketing easier. But as brands look at their tech stacks with AI, they are starting to scratch their heads and ask, “Why hasn’t AI solved my marketing problems yet?”
In an Epsilon survey, 93% of marketers said they’re planning to allocate at least 5% of their budget to AI initiatives. But, according to a new report from IBM that surveyed 2,000 CEOs around the world earlier this year, only 25% of AI initiatives have delivered expected return on investment (ROI).
So, what will push ROI for marketers? Data clean rooms. Why? Because harmonized, enriched data at scale gives AI the best chance to work effectively and accurately. This means that brands need to look beyond the limits of their own data. This will inevitably require a data clean room strategy.
For most people, the word “AI” is associated with generative AI that creates new content, such as images or videos based on patterns learned from existing data. It works because the Internet provides the Large Language Models (LLMs) with enough content to understand what the user wants. Creating an image of a flying dog is a rather simple task because the Internet is full of images of dogs and flying birds.
Marketing is not so simple. Most AI applications for martech/adtech requires predictive AI. This type of AI works somewhat differently: it’s looking for patterns across datasets to inform marketing strategies. The result is a prediction of future performance. When applied correctly, the AI can autonomously decide the best way to reach an outcome: who to talk to, where, and when.
AI needs tons of data to learn and make the right predictions at scale. However, artificial intelligence can only learn from observation. An AI agent can’t create a media plan optimized for efficiency if it can’t see the sources of performance. A recent Epsilon survey shows 49% of respondents said they’re concerned that model accuracy is affecting efficacy. Human intelligence is observing that artificial intelligence isn’t training on the right volume or quality of data. This is why the industry is looking at cloud-based data clean rooms to get the models what they need.
Data scientists, working on marketing problems, were early adopters of storage-based, access-controlled data clean rooms. These data environments took their name from the particle-free clean rooms that scientists in the physical world used for experiments. The name is a disservice to the present-day technology. Clean rooms are not a tool to isolate data for experiments. They are the control plane for secure data collaboration. This is what marketers need for AI to finally deliver the ROI, a way for the LLMs to learn from larger datasets.
Advancements in cloud data warehouse infrastructure open the opportunity for AI’s marketing future. Clean rooms are now logical not physical, limiting data movement and enabling ephemeral federated queries. The leaders in clean room technology for advertising are building on top of the cloud infrastructure to deliver solutions by bundling data, identity, and AI powered user experiences. These solutions manage complex data use policies and enforce rigorous privacy requirements unique to advertising. Once the data is permissioned appropriately, these tools enable marketers to ask the kinds of questions they have always wanted answered. Who are my best customers? Where do I reach them? How do I optimize the ways I am reaching them?
There’s an old adage that applies to AI: garbage in, garbage out. AI is only as good as the data it’s trained on.
A brand’s owned first-party data (like names, addresses or emails) is incredibly useful when it comes to building marketing strategies, but it’s not very wide in scope, nor is it always accurate and complete. Even the largest online retailers, flush with emails, still need a way to understand their growth opportunity—their prospects—by working with partners. A solid data foundation needs enhancement from data inside and outside of an organization.
A clean room employing identity resolution can harmonize, cleanse, enhance and connect all these disparate data points to an individual, pseudonymized profile. The importance of identity resolution as a native function in the clean room application cannot be overlooked. A single identity spine gives brands the best view of their customers—what they’re doing outside of the brand’s limited scope—and create stronger models for who their prospective customers might be. A marketer with a clean room is empowered to look for data that can fill in the blind spots.
According to the IDC MarketScape: Worldwide Data Clean Room Technology for Advertising and Marketing Use Cases 2025 Vendor Assessment, clean rooms have evolved from a “nice to have” technology to a “must have,” largely because of their ability to work with data and their ability to navigate the next frontier of data collaboration for use cases like AI modeling.
The answers to the most valuable questions in adverting are still trapped in a myriad of data silos. Technology can’t solve the data fragmentation problem alone. Brands, publishers, retailers, and data aggregators need to look at their own data assets and ask themselves if they are ready to participate in this evolution.
Gartner reported earlier this year that, “despite an average spend of $1.9 million on GenAI initiatives in 2024, less than 30% of AI leaders report their CEOs are happy with AI investment return.” The unfortunate reality is that most companies simply don’t have the scale of quality data necessary for AI to help them. The companies that are flush with data are struggling with data consolidation and managing usage rights.
The good news for marketing leaders is that AI is not changing strategic imperatives. In fact, it is reinforcing them. Investing in a strong data foundation is still the most productive way to build towards marketing efficiency. With the regulatory landscape constantly in flux, building a trusted identity relationship with your current customers has never been more important. The next step is finding the right way to extract insights and opportunities from paid, earned, shared and owned data assets. That is why every company needs a data clean room strategy.
The data that AI requires to answer the most pressing marketing questions about prospective customers will never be shared freely. Data partnerships, programmatically managed by data clean rooms, are a hard requirement. These tools enable data on unknown prospects to be analyzed alongside data from loyal customers with proper usage and privacy controls. Companies that recognize this and put clean rooms at the center of their marketing technology investment will benefit most from the next wave of AI capabilities.
AI alone cannot solve your marketing problems, but with the right data-centered strategy supported by data clean room technology, AI will deliver a return on investment by answering questions that drive efficiency and new customer acquisition.
