AI CLOUD
J.D. Power has developed AI Cloud, a solution that enables the creation of a cloud-based data platform supporting advanced analytics and AI. It allows users to ingest, transform and aggregate data, apply analytics, visualization and machine learning (ML) model training to the data.
This solution as it exists today was built by engineering, with internal stakeholders as the primary users. This includes software engineers, data scientists and analysts.
UX Project Objective
Develop the user interface and user experience for the creation and modification of Projects, Workspaces and Datasets within the platform. This includes the process of adding datasets (data ingestion), data verification/adjustment and ongoing status of datasets that have been added/connected to a workspace. In effect, this will be version 3.0 of the user experience.
Meet with…
data scientists and data engineers to determine:
How are users using the AI Cloud platform today?
What are some bottlenecks using the AI Cloud?
What work-arounds have users developed to overcome difficulties?
What are the areas of improvement with AI Cloud?
How confident/proficient are users using the AI Cloud?
Methodologies
Contextual Inquiry
Observe users using the AI Cloud 2.0 and observe them navigate through the platform
User Interviews
Ask users to identify any bottle necks or areas of improvement. Collect honest feedback and suggestions
N= 10, (5 data scientists and 5 data engineers)
Key Findings
Current hierarchy of the platform is not intuitive and data is often lost or duplicated
The overall layout of the platform was not designed for scale, features and updates are often tacked on wherever it seems to fit
The lack of a clear organizational structure has led to users wasting time searching within different workspaces for relevant data
Lack of proper governance has led to the creation of unnecessary workspaces
Users require various workspaces for publishing data to production, testing machine learning models, and aggregating data
Users have created multiple duplicate workspaces for unique purposes leading to a large unmanaged collection of workspaces
Admin rights and ownership is not intuitive and causes issues when requesting and provisioning access to workspaces
Lack of organization is visible within workspaces as datasets are further nested within eachother
The horizontal cabinet style of navigation makes it difficult for users to access different datasets.
Various AI algorithms and machine learning models were nested within different groups
Child Spaces were created to try and better organize groupings but created more difficulty and caused more confusion
Different tools would force users into different views
Current layout does not allow for multi-tasking or the ability to quickly reference datasets
Users would work around this experience by opening multiple tabs in multiple windows to access information
Switching between data applications overtakes the experience and any work will be lost
See images below, when creating a custom query and switching to data pipelines, the experience is overtaken by the new selection and all current work will be lost
AIC Cloud 3.0
Organize the platform to better separate environments and tools
Cleary distinguish between projects & workspaces you are entitled to contribute to
Provide each user with a personal workspace for the purposes to testing without the need to create a new environment
Strategically separate the main AI Cloud platform from third-party applications and integrations
Creating an overarching group type called “Projects”
Group related workspaces to help consolidate related teams, products, and work streams into one location where assets can be shared
Display administrative details and implement new tools for governance
This grouping mitigates the nested workspaces and child workspaces that have been used as a work-around
Platform level integrations can be accessed in one central location
Showcase the suite of tools and integrations so users do not disrupt their most active workflows
Introduce unique options and management within each third-party application
Implement global navigation for a more streamlined workflow
Allow users to navigate between workspaces without having to travel back to the landing page
Showcase metadata for users to view most recent activity
Tools and options are now localized to the page level
Proper management tools and governance were implemented at the appropriate levels
Project and workspace level tools are presented as a fly-out
This format allows for users to remain within the experience and not lose their place or any previous work
Fly-outs also allow the ability to quickly reference resources without navigating away from the experience
Dual Canvas Workspace
Flexible canvas sizing
Allows for users to adjust their canvas size for flexible utility
Two sets of data tools
Two sets tools allow for users to mix and match tools for use
