Role & Responsibilities

 UX lead designer  2023 Q3-Q4
  • Led the Ai-driven 0 to 1 product development.
  •  Influenced strategic vision and roadmap.
  •  Drove and facilitated a workshop to identify core issues with key cross-functional stakeholders 
  • Increased awareness and visibility around our Nitrate team's work on Generative AI across all YouTube UX teams

Two focus areas to integrate generative AI capabilities

01. YouTube Creative Ad tool

By integrating AI functionalities into the creative ad platform used by the YouTube marketing teams, YouTube could automate the creative ad creation process, significantly saving time and resources. Previously, we relied on external creative agency partners to manually produce creative assets, including copy and images.

02. Nitrate* formats (YouTube utilizes Nitrate formats, which are in-app messages aimed at driving purchases of paid offerings)

By pairing the AI function with the targeting system, we could generate more personalized creatives that resonate more effectively with individual users, potentially driving higher engagement and acquisition rates for the paid product features.

The evolution of the Nitrate AI MVP project

Our MVP project was born from our Bard hackathon event in Q2. I was on the team that developed this idea.

Within less than two quarters, we were ready to run an experiment against the manual promotion to see if automated creative images would outperform or get neutral metrics in terms of CVR and CTR. We launched a premium upsell campaign by utilizing existing Google AI platforms such as PaLM for text-to-text generation and Miro for text-to-image generation.

Marketing has been identified as a top business function for which generative AI could deliver significant improvements.

According to a recent study by the well-known business consulting firm McKinsey & Company, four areas have been identified as accounting for approximately 75% of the total annual value from generative AI use cases, with marketing being one of the most promising among them. (These areas include customer operations, marketing and sales, software engineering, and research and development.)

The study also found that generative AI has the potential to boost marketing productivity by 5% to 15%.

There was a major opportunity for us to effectively streamline our team's workflow and enhance the marketing team's velocity through the power of AI.

Nitrate generative AI aspirational roadmap

I partnered with a product manager to prioritize ideas and define a roadmap, streamlining processes for our team and delivering value to cross-functional stakeholders.

As a person with full knowledge of both the implementation processes of an internal campaign generation tool and the front-end visual presentation, I brought insights and came up with ideas that could optimize our team's processes and benefit our cross-functional stakeholders. 

Nitrate generative AI workshop

01

Nitrate Playground

01. Enable AI input fields via toggle : Users activate the AI input field by simply toggling a button.

04. Fast follow-up features

Allow users to save output results or visit the prompt history.

Given the rapid evolution of AI technology, I organized, facilitated, and led a workshop with key cross-functional stakeholders to identify opportunities for future AI implementation. The workshop consisted of two sessions: one brainstorming ideas for utilizing the internal marketing platform to generate creative YouTube ads, and another focused on reimagining the future of Nitrate formats, the containers used to display creative images.

02. Text-to-text generation powered by PaLM

  • Prompt entry: When users activate the AI input field by simply toggling a button, an additional text field appears above the manual field.

  • Input selection: Users can select one generated option to preview the image on the right side. 

    The ability to like or save options will be included in future launch.

    Users can override the text by entering manual text if they don't like the suggestion.

  • Visual indicator during processing: A loading animation, designed to visually represent AI activity, is displayed for up to 12 seconds while the system processes prompts.

03. Text-to-image generation powered by MiRO

All images generated by MiRO are legally approved for use. As long as they do not display any human figures, they are ready to be used in marketing campaigns. Once we have finalized the images for testing, we will go through our usual campaign review process with the marketing, UX, and legal teams, just as we do with all other campaigns.

Although it takes about 12 seconds to load images, it is a significant improvement over the traditional way of manually creating background images by our agency partner, which can take up to 4 weeks.

Experiment results & GenAI Learnings

We launched an experiment using 4 AI-generated creative ad variations alongside a human-crafted control from an external agency. Despite technical limitations preventing the use of AI copy within character constraints, we only launched AI-powered creative images.

Our YouTube Premium plan upsell campaign using AI-powered visuals achieved neutral engagement, with some visuals even surpassing the human-made control.

02

Future of Nitrate formats

Overview 

To adopt the fast-changing industry's technology and empower our team's work, I led the design effort of transforming a hackathon idea into a feasible solution for the YouTube marketing team. This involved utilizing Generative AI capabilities to integrate into the internal marketing tool and efficiently generate AI-powered creative assets.