AI Builder
AI Builder is a turnkey solution that brings the power of AI through a point-and-click experience, so you don't need coding or data science skills to access the power of AI. With AI Builder, you can build custom models tailored to your needs, or choose a prebuilt model that is ready to use for many common business scenarios.
“"The G&J Pepsi IT team uses AI Builder and Power Apps to create the next generation of our Store Audit app, enabling better mobility and efficiency for our field personnel. AI Builder will help build an AI model to automatically identify and track our products using object detection. For our field worker, it will be as simple as taking a photo and letting AI Builder do the rest.”
Eric McKinney — Enterprise Business Systems Manager
G&J Pepsi
Role: UX Design
Design Team size: 1 UX designer to support Redmond team (me), 6 UX designers supporting Paris team, 1 Content designer
Time on product: 1 year
Responsibilities: Establishing triad relationship and design best practices, coaching product partners new to semester work style and working on user-facing features. Contributed heavily to product envisioning. Responsible for maintaining product cohesion across Redmond and Paris groups. Built partnership with data scientists who created the models, maintained parternship with CELA, and conducted day-to-day design feature and research work.
Sample feature: Prediction score re-design
How can we let users know how their models are performing without presuming their business needs or suggesting a destructive action?
Feature timeline: 4 weeks
Role: Design lead
Design team members: Worked with content designer, Design researcher and consulted with CELA
Feature context: user group
A key distinction for our product and this feature is that our core users do not code. Our product is designed with the understanding that our users have no framework for developer concepts and no experience with data-science or AI.
Existing experience
The original design shows a performance number in a ring (see left). We initiated this work as our research and customer interactions told us users did not understand this display of information. We were also wrestling with how to encourage users to publish and use their models.
By updating the performance score design we wanted users to feel more confident about the state of their model and deciding either to publish or update the model.
Sketching
User research
Conducting A,B,C User test
After brainstorming with Dev, PM, Science and Content design we set to work on preparing an A,B,C test. We wanted to pin down our users comfort level. What is the most simplified experience we could make? What would a very complex experience look like? And most importantly how much information did the user need to increase trust and confidence in their models.
Revisions
Research results indicated that while a simple letter grade representation was widely preferred by both novice and advanced users the information in the score’s detail panel demonstrated an opposing preference. Users loved the clarity of the simple ABC representation but when drilling down into what this means expected and enjoyed seeing the more comprehensive supporting data (even though the less technical users didn’t have a full understanding of each part).
Outcomes
Not only has this work increased the rate of publishing by several percent but this feature has also set a foundation for sibling products using the same technology. My close partnership with research, content design and the legal department has resulted in this feature being used as a case study to evangelize AI best practices across disciplines and the product suite.
Customer insights leveraging our design and research
Dynamics 365 Sales Insights encorperating our results