Generative AI in DSPs: Balancing Innovation with User-Centric Development

We're roughly two years into generative AI being available to the masses, and many of us are still looking for ways to grasp some of those so-often-promised productivity gains. During this time, generative AI has brought both excitement and practical challenges. While initial expectations promised rapid productivity boosts, the reality has been more nuanced, involving complexities in integrating AI into our daily workflows.

Still, generative AI has opened up creative possibilities that were previously out of reach, empowering users to explore new ways of working, however that looks for them.

Is it Generative AI or AI/ML?

It's worth clarifying what I mean by generative AI, as it differs in key ways from more traditional artificial intelligence and machine learning (AI/ML). At the risk of oversimplifying too much, we can distinguish them by their general uses:

• Traditional AI/ML: Excels at predictive tasks by utilizing lots of data (e.g. bid optimization, lookalike modeling).
• Generative AI: Generates content and carries out tasks based on natural language input from users.

The accessibility and popularity of tools like ChatGPT and Claude have redefined what everyday folks mean when they say they want or expect AI capabilities in a platform. We know that products use AI/ML and treat that as table stakes; now, we want products to interact with us on a more human level.

In the context of demand side platforms (DSPs), AI/ML continues to be the breadwinner for performance gains and is a key differentiator among ad platforms. A few examples: AI/ML ensures bids are efficient and effective, improves audience  targeting, and generates recommendations for users that lead to increased ROI. This is exactly what Yahoo Blueprint Performance has achieved. We’ve recently shared some of the impactful results advertisers are seeing by leveraging this type of AI/ML.

For these platforms, generative AI presents a broad opportunity to accelerate user workflows and increase the usefulness of our products. We're already seeing experimentation and execution in areas such as:

• Campaign ideation. Assisting users in brainstorming campaign concepts based on their goals.
• Ad copy generation. Providing creative ad suggestions tailored to specific audiences.
• Workflow automation. Simplifying complex processes through conversational interactions.

However, losing sight of the user can lead to expensive features that are underutilized or even ignored. For example, a generative AI feature that outputs too much text in response to a simple question can overwhelm the user, ultimately discouraging them from using the feature anymore.

Avoiding POMO

In a hyper-competitive space, products (or services) are often compared to one another, and it's natural for feedback to focus on what one product has that another product lacks. When new technologies are involved, this can result in a perception of missing out (POMO); that is, feeling like the longer you take to adopt the latest tech, the more behind you become. The risk, of course, is that you succumb to this pressure and abandon your own principles, chasing features that ultimately may not solve the right problems.

A User-Centric Approach

Whether building a product or providing a service, it remains critically important to stay focused on the end user. Generative AI shouldn't just be about adding flashy features; it should be about enhancing the user experience in meaningful ways. A user-centric approach means involving users early in the design process, gathering their feedback throughout, and iterating based on their needs.

By understanding user needs and aligning generative AI features to solve real pain points - like reducing time spent on repetitive tasks or providing clearer insights - we can avoid the pitfalls of building tech for tech's sake.

This approach ensures that the features we create are genuinely useful and seamlessly integrated into user workflows, ultimately driving adoption and satisfaction.

How Yahoo DSP is Integrating Generative AI

Here at Yahoo DSP, we're not sleeping on generative AI; we're putting the time in to ensure we pursue features that will genuinely uplevel our users. We're not going to leave you empty handed in the meantime, though, and we're excited to announce that starting now, our users have access to Blueprint IQ! Our assistant is a DSP expert. It's ready to answer all your help-related questions, saving you time by getting you the information you need so you can finish your tasks faster and with ease.

Log in and start using Blueprint IQ today!