Omnicom's AI chief shares what Madison Avenue knows about you.

How AI is changing the ad game, according to Paolo Yuvienco.

Paolo Yuvienco sits on top of a giant data set of information about the consumer. As chief technology officer of Omnicom his job is to figure out how to supercharge it all with artificial intelligence and keep the ad business ahead of the game. 

He joined the company in 2017 from Publicis’ DigitasLBi and in this conversation he gives us a peek inside the black box, detailing how AI is helping big ad companies - and their big tech partners - unearth much deeper insights into our buying patterns.

Q: Tell me something that’s been an “aha!” moment, something you couldn’t see before but now with AI you can?

A: The “Aha” was probably early 2022 when we started to see OpenAI models surface interesting work they were doing. We knew immediately that this was something that was going to affect not just our industry but every industry. You can go back to 2017 when Google was trying to effectively map how the human mind works. 

Omni is our marketing operating system. There are a bunch of apps that sit on top of it. The foundation is this massive data set - probably the largest dataset in the industry. And it consists of first, second and third party data that’s being sourced, compiled and matched in a way that gives us insights around individuals and their behaviors, their demographics so think website behaviors, what sites people visit, their purchase habits, what they’re buying, all different types of data that we’re sourcing and compiling inside of this ecosystem. 

Then there’s apps that sit on top for audience building. I need to sell 50,000 new vehicles in North America next year. We need a campaign that’s going to launch against this new vehicle launch, identifying the audience to target and tools to derive a media plan, an investment planning pool, and a cultural analysis tool. It could structure the unstructured nature of culture with overlays of behavioral data. You know what people are doing with the culture app and you can almost predict why people are doing it so it gives us a much better context to start building better audiences, better media plans and better creative assets. 

When we took the LLM, we were the second company in the world, after Disney,  to get access to the OpenAI model in Microsoft’s Azure. We’ve had access to the OpenAI GPT models since January of last year.  The ability to create human conversation with audiences, that was a real revelation. 

Being able to surface those insights and being able to understand edge cases of audiences that we’re building for specific clients and campaigns – that was a bit of a revelation. So early last year when we started integrating the LLMs into Omni that’s when it really started clicking that this is going to massively transform our business. 

Q: Help me understand something more concrete?

A: Insights are so much faster. If I’m opening a clothing line, I start building audiences against who I’m competing against, and I want to understand a bit more about those audiences so I might ask; ‘Can you surface some of the non-obvious websites that this specific audience [visits] so I can better target?’

Those edge cases would take weeks to mine that data and now I have that audience, helping me identify the top five influencers that I can use in my campaign. It would be able to identify that now, and it still needs to be validated by humans but it understands the outcome we are trying to drive.

Here’s why that’s a significant leap, because now we can be far more strategic in how we approach things whether it’s creating your plan, media plan, activation plan or content creation and content adaptation. So scaling content, we think about mass personalization at scale that becomes a reality with the implementation of generative AI. Today there’s not enough processing power to do that in real time but think of the day when every ad shows a specific individual, where that individual resides, it puts an auto ad with a background in their neighborhood. It knows that they have a family unit of four and it puts four people in the vehicle. These types of personalized engagements become possible with the ability to start generating on the fly. 

Q: Are you building your own AI models or partnering with OpenAI? Can ad agencies be competitors to big tech companies? 

A: I can foresee a day where we might have our own foundational models but that’s certainly not the plan today. We’re working closely with OpenAI through Microsoft and Google Vertex AI environment and Gemini. We’re working with Amazon with their Bedrock platform; Adobe, Firefly, Getty, Shutterstock. We’ve struck partnerships with all the major players so we can work collaboratively.

Q: What are the things slowing progress down?

A: The legal issues are certainly at the forefront of this. It’ll get sorted out either in the courts or through the industry. I don’t think it’s going to be a massive barrier in the future. I think ownership, IP, is going to be more of a debate. What happens when humans are partnering with machines for creation? I think there’s still some questions around that. Image and video, that’s going to be something that we’re still navigating.  

Q: What is your role in terms of overseeing AI at Omnicom ? 

A: The first leg is around large multi agency global pitches. When I joined Omnicom, it was under the assumption that I would help define and operationalize the transformation agenda for large clients. What it also translates to is doing the same for our agencies. It’s working closely with their tech leaders to drive a more holistic strategy that can bubble up into an overall client solution.

The second leg is around our platform strategy across the group. You’ve probably heard of Omni, our operating marketing system. We just acquired Flywheel that’s focused primarily on commerce and transactions and then there are various other platforms…some of which are being driven by Adobe products.  

The third leg is a bit more nebulous: its strategic initiatives like offshoring; how do we create a more globally distributed delivery capability ? How do we automate, and do we use Microsoft for efficiency. We’re doing a lot of content automation at scale. 

AI has been part of our industry for well over a decade. We’ve used it to get predictive modeling around the work we’re doing for our clients. We’ve started to apply AI and machine learning for things like creative strategy, things that LLMs are really starting to make accessible. 

Q: Tell us about the steering committees that have been set-up? 

A: We wanted to make sure that we were approaching AI in an accelerated way but in a very responsible way so we set up steering committees which oversaw several work streams and sub committees. Responsible AI was one of the working groups - what’s it going to look like 3-5 years from now. 

We have many partnerships with media, publishers, tech. How is that going to change? So we set up roughly nine  different work streams. Each one of them was around what steps we need to take now to ensure we can prepare ourselves for the future? Things change almost daily. 

Q: How are companies saving money because of AI implementation?

A: One is having more intelligent automation that is definitely driving efficiencies and even content automation. Many clients are looking at how to modernize production capabilities. All of this still needs human intervention.

Q: What are the negative sides of AI that you worry about? 

A: I have concerns around workforce, in the sense that like every technological advancement, we know that it actually creates more jobs, so I’m not worried about reduction of the workforce but I want to make sure that we’re being mindful of not leaving anyone behind.How do we ensure that the technology doesn’t move faster than our ability to retrain our workforce. 

Q: How are you using AI in your own life? 

A: I’ve been able to create art. I’m a terrible artist, I think in ones and zeroes, but what I can do now is I can create beautiful presentations as opposed to presentations that were bullet points and images without having to scour Google Images.