How Agencies Are Using AI For Ad Data Analysis

Hello everyone, and thanks for joining us today. If we haven’t met, I’m David Burkwit, the founder of the AI Marketers Guild, and I’m here with John Riley, co-founder and CEO of AIO. I’m excited to learn more from John today about how agencies are using AI for data analysis. I’ll hand it over to him to share insights on how AI is changing the way we use data. Please share any questions or comments in the chat, raise your hand if you want to ask something live, and thanks for being here with us.

**(00:49)**  

Thanks, David, and hi, everyone. I’m excited to be here and talk about the data side of AI and some machine learning applications as well. Both generative AI and ML are having a huge impact on how marketers use data for decision-making and analysis, especially in the ad space. I’ll walk you through some of what these technologies make possible and also share a bit about what AIO does.  

**(01:25)**  

A bit about my background: I started out as an electrical engineer designing televisions at Sony, back when CRTs were still around. Most of what I know now comes from product management work, especially at Sonos, where I managed many wireless audio products. I later moved to a 3D printing startup, Markforged, where we created carbon fiber and metal parts. During my time there, I ended up overseeing marketing because we had issues with our sales pipeline. We’d stopped using Facebook ads due to low-quality leads, but then our sales slowed down. It turned out some good leads were getting mixed in with the noise, and we didn’t have a good way to filter and find those valuable leads. That sparked the idea for AIO.

**(03:37)**  

At the time, there weren’t any tools that people without specialized training could use to make better data-driven decisions. While there were ML solutions, they were mostly offered as services rather than self-service tools. So, we set out to create a self-service platform for data analysis. Over time, as AI tech evolved, so did our platform, making it easier for anyone to work with data—without needing advanced skills.

**(04:11)**  

Today, AI is everywhere, and there’s a lot of hype around it. That moment when people could suddenly use a simple chat interface to interact with AI—like with GPT—changed everything. Now, with AI, you can generate content, images, video, and blog outlines, which has transformed the creative process. But it goes beyond content creation. Clients increasingly expect data sophistication from their agencies, and large agencies and holding companies are investing heavily in AI.  

**(05:44)**  

When we first started, our idea was to help internal teams make data-driven decisions. But it turns out agencies are leading the way here. Agencies are early adopters of this technology because they can deliver a higher level of sophistication than their clients can. CMOs, in particular, are under a lot of pressure to prove that marketing investments are working and increasingly rely on data-driven methods.

**(07:20)**  

The old way of doing things was resource-intensive. Marketing teams would depend on data scientists or analysts, but those teams are often focused on the company’s product or core technology rather than marketing. For example, at Markforged, our data scientists were focused on perfecting 3D-printed parts, not marketing. Working with external analysts often creates delays and miscommunication, which leads to time lags and missed opportunities.  

**(08:44)**  

Now, with AIO, you can ask data questions directly and get answers without relying on outside analysts. You can even share a data analysis interface with clients, allowing them to self-serve for straightforward questions. This kind of access helps optimize ad spend, enables predictive modeling, and allows faster decision-making.

**(09:52)**  

Our approach with AIO is to make working with data as easy as possible. Agencies can White Label the platform and build custom models and dashboards, merging client data with their own proprietary data if they wish. This setup allows agencies to create unique audience segments and identify factors that contribute to purchasing.

**(11:26)**  

Here’s a quick example from our customers. Zenith is using AIO for a next-gen business intelligence approach, where users can simply chat to get answers to data questions. Fathom uses it for predictive enrollment targeting. Building dashboards, which used to take a lot of time, can now be done on the fly, allowing agencies to provide insights to clients faster. We’re even working on templates so that once a report is set up, you won’t need to rebuild it each time.

**(12:25)**  

For the demo today, I’ll show you how AIO helps analyze campaign performance data. This is a common scenario where agencies need to assess past campaigns to plan and optimize future ones. Many agencies even use AIO’s data chat as part of their pitch process by quickly analyzing a client’s past ad performance.

**(13:01)**  

David, you mentioned that the idea of certifications is fading out. That’s exactly what we’re seeing. Today, when only one person knows how to use a tool, it creates dependency issues, especially when they leave. With AIO, you don’t need certification or training to work with data. You can connect the data, and then the language model will automatically interpret and run queries for you, whether in SQL or Python.

**(14:12)**  

Our platform can learn from past data patterns and predict future outcomes. For example, if you’re lead-scoring, AIO can help predict lead conversion likelihood, allowing you to assess your channels’ performance in real time by looking at lead quality. The more data features—like demographics or customer attributes—that you feed into the model, the more accurate it becomes.

**(16:22)**  

We designed AIO to work with any type of data, even publicly accessible or social media data. For instance, we have a sentiment model trained on millions of tweets, labeled based on positive or negative emojis. This can be used to monitor social media sentiment in real time. You can connect virtually any dataset, whether that’s marketing data, CRM, or social feeds, and analyze it together with other data sources.

**(18:47)**  

Let’s go through a sample data chat session with AIO. Here’s some campaign data where we ask for an overview, and AIO calculates impressions, click-through rates, costs, and more. After generating results, AIO also explains the method it used to produce the answer, which helps you understand the process behind each insight.

**(19:16)**  

We can dive deeper, for instance, by asking which metrics varied over time. AIO found that cost per click had a high degree of variation and was worth investigating further. It can also break down metrics by factors like campaign type, placement, and time period, generating charts for easy analysis. If you’re trying to optimize cost per click, AIO can even suggest the best placements or campaign types to focus on based on your data.

**(21:38)**  

For reporting, AIO can also generate summaries. You can ask it to draft an email to a client with embedded data insights, making it easier to share results directly.

**(22:20)**  

Yes, you can customize charts, including colors and branding. We’re also launching a feature that allows live editing of charts to match your brand colors.

**(23:26)**  

AIO is a White Label solution, so you can customize it for your agency’s brand or for specific clients. Clients can self-serve data insights or use the reports you create for them. You can even guide them through chat history, embedding past questions to lead the analysis.

**(27:48)**  

To ensure the accuracy of responses, we use a mix of GPT-3.5 and GPT-4, choosing models based on the complexity of the question. We also have a validation mechanism where responses are evaluated by OpenAI’s vision model to confirm that they correctly answer the user’s question. This setup lets us adapt as new, more efficient language models become available.

**(30:21)**  

To get started in AIO, you can create a new project and connect your data from sources like Google Ads, Snowflake, or CRMs. We support role-based access, so each client’s data remains in a separate team compartment. Once data is uploaded, AIO identifies the data types, runs column correlations, and offers options for data cleanup and transformation, such as removing outliers.

**(33:45)**  

You can also merge data sources. After data is merged, you can start exploring insights or sharing the chat interface with clients. You can customize the interface with client-specific instructions, suggested questions, and more.

**(35:11)**  

If you want to predict outcomes, AIO includes a predictive engine. You can build ML models for time series, classifications, or numerical predictions. For example, you can create a model to predict return on ad spend based on past data and key factors. AIO’s machine learning features make it easy to identify data patterns and provide actionable predictions.

**(36:34)**  

After building a model, you can deploy it in real time for ongoing predictions, such as scoring new leads or optimizing ad spend. This lets you apply data insights instantly, which is especially useful in dynamic ad environments.

**(38:04)**  

We often get questions about data access. Agencies are usually able to get marketing data access from clients, but getting revenue data can be more challenging. Sometimes we work with high, medium, or low value categories instead of exact revenue amounts, which still provides valuable insights for the models.

**(41:10)**  

Marketing mix modeling is possible with AIO,

 as long as you have a data set that includes spend data across different channels. By comparing results at different spend levels, you can optimize ad allocations. We’re working on scenario modeling features to make it even easier to adjust spend based on outcome predictions.

**(44:35)**  

AIO makes it possible for anyone to explore data, and some data scientists appreciate this because it frees them from routine requests, allowing them to focus on high-value tasks. We find that our platform often helps educate teams on data, leading to better questions and analysis.

**(47:32)**  

Currently, AIO works on one dataset at a time, though you can merge additional data sources. In the future, we aim to allow for cross-table queries using a knowledge graph structure, which would enable dynamic querying across multiple datasets.

**(50:48)**  

Getting started with AIO is simple. You can log in, connect your data, and start analyzing within minutes. Our team is here to support and guide you through any setup or customization.

**(53:09)**  

Thanks, everyone, for joining today, and a big thanks to John for the insights into AIO. This was incredibly helpful, and we’ll share the recording and additional resources for following up. John, thank you for spending time with us and showing what’s possible with AI-driven data analysis.