Few can genuinely claim expertise in implementing AI today, leaving the landscape as ever-changing as the Wild West. To demystify AI marketing tools – their applications, benefits, and more – FutureWeek is tapping into the insights of tech and marketing leaders. This week, we spoke with Anna Calabrese, VP of Data Science and Engineering, at supply-side platform OpenX. We find out how her leadership of the company’s 40-person data team has helped to deliver advanced AI-powered innovations that elevate the performance of the ad exchange, and more.
Can you describe how the OpenX tool works? Particularly in relation to AI?
Results by OpenX is a suite of approachable AI capabilities that harness deep neural networks (DNNs) to curate high-value inventory to deliver against specific advertiser goals. Our models examine the stream of requests the ad exchange receives and learns the signals that are indicative of outcomes. The model uses these learnings to generate an array of probabilities in response to new requests, predicting which ones are highly likely to lead to an outcome that a certain buyer desires, and route the traffic to that buyer. This means we are only sending out opportunities to the buyer that have a high probability of driving the results they want.
Which AI use case are you most proud of so far?
Our recent focus has been on developing and testing solutions that identify and prioritise the highest-value inventory to bid on based on buyers’ desired outcomes. Having the ability to identify the best-suited media placements and automatically send only inventory that closely matches the buyer’s campaign objectives — while taking behavioural and other audience signals into account — is a game-changer for the digital advertising ecosystem. I’m immensely proud of these solutions.
Who in media and marketing do you think stands to benefit the most from the AI revolution?
There are possibilities for just about every company in the ecosystem to benefit from AI, but the greatest rewards will be for the companies that have the right data foundations to embrace these new tools. Having the right data foundation is a crucial pain point for many companies when trying to build AI applications. Most companies hit roadblocks not because the models are bad, but because their data foundations are lacking. Three of the core problems companies typically face are poor data quality, data silos, and a lack of metadata and data lineage.
Which parts of the media and marketing supply chain are exposed to the highest risk of disruption in your opinion?
There are opportunities at virtually every point in the supply chain. From the application of highly efficient AI to curate inventory and surface only the ad opportunities that are likely to drive value for buyers to automated campaign management enabled by AI agents that constantly examine data to optimise campaigns, we are innovating to drive better results across the digital supply chain. Looking beyond that, AI-generated creatives with a human perspective have the future potential to power greater personalisation but also elicit the desired sentiment and generate deeper engagement with the user. In these areas, there is still a lot of development required, but the potential is undoubtedly interesting.
Of all the societal challenges associated with AI, which concerns you the most and why?
For me, the biggest potential concern is bias in AI. If AI models are built by teams from homogenous demographic groups, they’re going to exclude a lot of perspectives and therefore people. Diversity of thought is required to build AI systems that are truly inclusive and representative of all the people they affect and are designed to serve.
Have you made changes to your organisational set up for both AI adoption and to keep pace with the technological evolution?
OpenX has leaned into AI for a while now. We’ve long been enabled by being cloud native, and we’ve been using AI efficiently for many years to optimise aspects of our business like flooring, fee optimisation, and curation. More recently, we have created a team that evaluates various AI tools and how they can best serve not only the business but more importantly provide the most benefit to our employees in their roles.
What has been the biggest challenge in integrating AI internally?
I think the challenges we’ve experienced are the same challenges as most organisations – poor data quality, data silos, and a lack of metadata and data lineage.
Data quality: Think “garbage in, garbage out.” Inconsistencies, missing values, and outdated records are a big challenge in successfully integrating AI. Even the best models can’t overcome bad input data.
Data silos: Across organisations, it’s more common than not that data is scattered across teams, tools, or systems that don’t talk to each other. This makes it difficult to build a unified view of a business, system, or user.
Lack of metadata and data lineage: This occurs when users don’t know where data came from, how it’s been transformed, or how reliable it is. Without trust in data lineage, teams are hesitant to use data in high-stakes AI applications.
What will be the net impact of AI on employment in your opinion?
Like any other revolutionary technology, AI will ultimately impact numerous job functions across just about every industry. AI will help many people perform better in their roles by giving them immediate access to the information and data points they need to make decisions while removing repetitive, time-consuming tasks from their to-do lists. While every company will make different evaluations, at OpenX, we view AI as a tool, not a replacement: AI and machine learning require significant human input to be effective. The quality of our solutions is driven by the talent and creativity of the teams behind them.
What about the AI future are you most excited about?
AI is going to help the digital advertising industry enter a new era where media partners, advertisers, and consumers all benefit. We are doing our part by developing approachable solutions that can help buyers streamline operations, improve performance, and deliver more efficient ad spend without using personal identifiers. In doing so, we are helping generate incremental revenue for publishers while delivering relevant ad experiences for consumers without compromising their privacy – and that’s really exciting to me.
