Supply Chain Agent

The business wanted to speed up searching for assets because the current system had a very complicated search tool that created a bottleneck.

Affects: Ad Sales, Advertising, Entire supply chain

My role: Lead UX Designer/Researcher

Timeframe: 4 weeks

End Users: Ad Sales, Content Researchers

Process:

  • 5 user interviews

  • Competitive Research

  • Compliance Testing

Goal

The product owner wanted to use a Generative AI chatbot. NBCU already had an LLM (Large language model) called Prism Portal that we could use as a model and add on the videoplayer.

We wanted the users to be able to have a “conversation” with the AI to quickly find what they needed.

e.g.: An ad sales user may want to collect clips to make a reel for “Family Night”

Research - Old System

The users searched on a very complicated filter drawer to find assets.
First, they had to choose from 10 categories, and within those categories, there were over 80 fields from which they had to describe what they were looking for.

Process

I explored how we could offer the chatbot to users while they used the current database.

I sketched out 6 ideas for how this could work.

⭐ We went with the free-standing, full-screen agent. From here, users could begin their search.

First version

I adapted the interface to incorporate the third-party video player to display its description to the left of the video.
This is with the information that I had up to this point.

Pivot:

The PM realized that both the 1) video and 2) its tagged scenes (when someone says “Family”) each have their own descriptions.

I needed to design a way to display both but make it clear which was which.

I also wanted to incorporate copying the whole list of videos, a partial list, and the timecode.

Solution

I divided the page into chunks:

  1. Video list

  2. Video player

  3. Video description and metadata

  4. Tagged scenes and their descriptions

When the user finds what they need, they can export the list to the main supply chain app and order them.

This is the end of the happy path.

Result

I delivered Figma files, annotated userflows, prototypes, and the Prism design system.

We tested this flow with 5 user groups and they found the system to be faster and more efficient than the current method of searching.

Much of our results are measured by Pendo.

Start Page

Response

Side panel

Actual productivity and accuracy are projected to increase by 66%.

Final Thoughts

The dev squad is still working on this product. I’ve done two rounds of QA for functionality and design and I’m sure there will be more.

I look forward to more AI projects in the future now that we’ve gotten started with the two products that we worked on.

As we move forward, I realize that we need some guardrails for the AI bots before it’s too late.


I’m working on documents for the UX team and to propose for NBCU regarding a framework that we can use when creating a new solution involving AI.