Fabric: AI Order Cloud

Fabric:
AI Order Cloud

Overview

The goal of this project was to seamlessly integrate AI into Fabric's eCommerce platform to align with the company’s strategic goal of automating key processes, improving efficiency, and enhancing user experience for operational teams.

from Ambiguity to Alignment

I organized a workshop with executives, department heads, and key stakeholders, to align on the vision for AI integration. These discussions played a pivotal role in shaping the strategic direction and gave us (design) a seat at the table.

Identified Opportunities
  • Improve operational efficiency

  • Provide easily accessible analytics and performance data to non-technical people

  • Eliminate tedious and/or time-consuming tasks

Guiding Principles

I knew that, as a design team, we had a unique opportunity to not only influence how AI would look, but how users would interact with it. I pulled together a quick off-site in SF so we could strategize in-person as a team. I was very happy with the vision we came up with:

Bold

AI needs to be exponentially better than the current experience. It can’t just be “better”, it has to be amazing.

Trustworthy

Transparency is key here - AI should give a rationale for recommendations and provide an easy way to opt in or out.

Relevant

AI should help me do MY job faster by providing relevant, contextual information at the right time.

Unobtrusive

AI shouldn’t distract, obstruct necessary information, or derail tasks by requiring me to navigate to another screen to complete or verify an AI action.

Predictable

Use familiar patterns with AI interactions to match mental models and reduce cognitive load. If there is friction in the experience, it will not be adopted.

Consistent

AI interactions should be instantly recognizable as such (visually differentiated from other content), and entry points should be consistent across the platform.

Target Personas

After analyzing the results of our workshop, we focused in on two key personas who we felt would see the fastest benefit from automated workflows. We then interviewed a handful of clients matching that persona, this allowed us to narrow in on the actual workflows we might improve.

Merchandising Managers

Potential workflows:

  • Maintain product data accuracy & categorization

  • Analyze performance metrics

  • Manage inventory

Supply Chain / Logistics Manager

Potential workflows

  • Process orders via ERP/WMS systems

  • Ensure accurate order fulfillment

  • Optimize fulfillment operations

  • Track performance metrics

Team Brain

One of the most enjoyable parts about this entire project was the opportunity for the team to work closely together. We had several live sessions to sketch and revise potential layouts and interactions. We then each created a quick wireframe of our proposed workflow.
Not only was it a ton of fun, but the ideas that came out of it were exceptional (in my biased opinion).

Early Prototypes

Its a bit unusual to do prototypes this early on, but we needed to make sure that our CEO was onboard. He had initially wanted a "bolt-on" chatbot type experience, but I was able to convince him that we needed to go deeper if we wanted to really add value and gain market differentiation. Here are a few of the goals we had with our designs at this point:

1.

Building trust in the system by providing an easy way to for user to see why the AI provided the results it did.

2.

Interactions and transitions that felt natural and intuitive. No "WTF?" moments.

3.

Affordances for moving between the chat window and work page, so users didn't feel lost or stuck.

4.

Visual cues that clearly delineate AI content from user-generated content.

AI-Specific components

We next turned our attention to the UI design. The key here was to differentiate AI-generated content from user-generated content. We began with a simple style-guide and refined from there, eventually creating components that were then added to our exisiting design system.

Customer Sentiment

As we began to zero in on final designs and interactions we ran tests with existing customers to gauge the perceived utility of what we were building. In this test, we asked subjects to complete a pre-defined task using our existing UI. We then asked a group to complete the same task with our existing UI + ChatGPT. We asked the final group to use our new AI-enhanced workflow. The results were positive and gave us confidence in our direction.

As we began to zero in on final designs and interactions we ran tests with existing customers to gauge the perceived utility of what we were building. In this test, we asked subjects to complete a pre-defined task using our existing UI. We then asked a group to complete the same task with our existing UI + ChatGPT. We asked the final group to use our new AI-enhanced workflow. The results were positive and gave us confidence in our direction.

Industry Feedback

"I haven't seen anything like what Fabric is doing with their approach to AI. Real-time inventory will be game-changer for us."

PedMeds CTO

"I would put fabric OMS in the top quadrant for AI feature in an OMS. No one else is doing this."

Gartner Analyst

"We needed a platform that could accommodate everything – order tracking, delivery options, inventory management, etc. Now, with Order Cloud, what used to take months now takes days or weeks.

VP, Commerce Engineering, Chico's

Split Shipments prototype

Shipments fulfilled from two or more locations cost business money in the form of added shipping fees and handling time. Supply chain managers now had critical data instantly at their fingertips and could update fulfillment rules in minutes instead of weeks.

Description Analyzer prototype

Merchandising managers used to have to comb through spreadsheets, check analytics (often requiring additional support) to get even a fuzzy picture of how their product descriptions are performing. With AI they can now get a score on their entire catalog, looking at key words and SEO performance, in minutes.

Results

The results from our beta launch were very encouraging and a testament to the strategic alignment between our design decisions and Fabric’s business goals, demonstrating the power of a well-executed UX strategy.

49

49

%

%

Data analysis time was improved by 49% – a significant time savings

Data analysis time was improved by 49% – a significant time savings

90

90

%

%

AI results scored 90% accuracy rating in our beta launch

AI results scored 90% accuracy rating in our beta launch

40

40

%

%

Self-resolution rates saw a 40% improvement

Self-resolution rates saw a 40% improvement