A.I. Assist

Summary: The organization needed to incorporate artificial intelligence (AI) into a web application that tracked the movement of parts through autoclave machines. This tracking would reduce errors made while using this machines, which can cost over $100k. The AI needed to assist users in determining which parts should be processed first and indicate when multiple parts could be processed together.

  • Me and a group of my coworkers traveled to a plant in Riverside, California that uses autoclaves and the Autoclave Loading Tool. The Autoclave Loading Tool is the application that we would be adding AI to.

    While in Riverside, I started my contextual research. I walked around the shop floor and studied the workflow to understand how parts were being moved from other machines to the autoclave and see how workers were interacting.

    This was followed by 1:1 interviews with the users of the Autoclave Loading Tool. I aimed to gather information on pain-points and understand whether adding a new feature would be welcomed or require top-down change management. During this journey, I learned that the standard workflow occasionally deviated. This information was vital because the Autoclave Loading Tool would need to be overridden at times.

  • It all begins with an idea. Maybe you want to launch a business. Maybe you want to turn a hobby into something more.

  • With the autoclave workflow now known, I could start my analysis on the Autoclave Loading Tool, I mapped out every action and pathway that could be taken to achieve a specific goal. My analysis indicated that new features were being built in a ad-hoc manner. It appeared that no clear Information Architecture had been created prior to development. I knew that I needed to find a way to introduce AI into the application without increasing cognitive load.

    The last task I did before building a design was to analyze the existing data that the application was driven by. I wrote multiple SQL queries to ensure that I understood how the data was connected and that all new features added would not disrupt the existing data models. During this analysis, I also saw that there were a number of parts that would return null values due to not having dimension data. This made me realize that I would need to find a way for users to enter dimension data to some parts in order for the AI to work.

  • With all of the knowledge I had gained during my research phases, I was now ready to start building mockups. I had learned during the process that there were now two primary functions that the new feature required. The ability to seamlessly include AI and the ability to enter dimension data to some parts.

    Since entering dimension data would not be used very frequently, I placed it in a location where users already went to add new parts. This created a mental-model that was based on admin tasks. I included a toggle switch, which would only display dimension data when activated.

    The AI component of the application was placed on a page that was used to load parts. A button labeled ‘optimize batch’ was added. When someone clicked on this button, the form would autofill according to AI results. The page provided a new affordance, without any disruption to existing users. Clicking on the button was optional and the autofill gave users an incentive to use the autoclaves in an efficient manner.

  • AI was added to the Autoclave Loading Tool seamlessly. Our planning and research paid off. No rework was needed, usability increased, and time to develop was shorter than if research and design had not been included prior to the build. We have heard reports that it is reducing the amount of time needed to select which parts should run together. No additional training has been requested, indicating a high rate of usability. We are collecting large amount of data on this application and will use it to compare efficiency and error rates. The product has provided business value to the company, reduced time to complete tasks, and has not placed any additional mental burden on users.

Research

Very Large Autoclave - Source ASC Process Systems

Large Autoclave - ASC Process Systems

Me and a group of my coworkers traveled to a plant in Riverside, California that uses autoclaves and the Autoclave Loading Tool. The Autoclave Loading Tool is the application that we would be adding AI to.

While in Riverside, I started my contextual research. I walked around the shop floor and studied the workflow to understand how parts were being moved from other machines to the autoclave and see how workers were interacting.

This was followed by 1:1 interviews with the users of the Autoclave Loading Tool. I aimed to gather information on pain-points and understand whether adding a new feature would be welcomed or require top-down change management. During this journey, I learned that the standard workflow occasionally deviated. This information was vital because the Autoclave Loading Tool would need to be overridden at times.

Application Analysis

With the autoclave workflow now known, I could start my analysis on the Autoclave Loading Tool, I mapped out every action and pathway that could be taken to achieve a specific goal. My analysis indicated that new features were being built in a ad-hoc manner. It appeared that no clear Information Architecture had been created prior to development. I knew that I needed to find a way to introduce AI into the application without increasing cognitive load.

The last task I did before building a design was to analyze the existing data that the application was driven by. I wrote multiple SQL queries to ensure that I understood how the data was connected and that all new features added would not disrupt the existing data models. During this analysis, I also saw that there were a number of parts that would return null values due to not having dimension data. This made me realize that I would need to find a way for users to enter dimension data to some parts in order for the AI to work.

UX Design

With all of the knowledge I had gained during my research phases, I was now ready to start building mockups. I had learned during the process that there were now two primary functions that the new feature required. The ability to seamlessly include AI and the ability to enter dimension data to some parts.

Since entering dimension data would not be used very frequently, I placed it in a location where users already went to add new parts. This created a mental-model that was based on admin tasks. I included a toggle switch, which would only display dimension data when activated.

The AI component of the application was placed on a page that was used to load parts. A button labeled ‘optimize batch’ was added. When someone clicked on this button, the form would autofill according to AI results. The page provided a new affordance, without any disruption to existing users. Clicking on the button was optional and the autofill gave users an incentive to use the autoclaves in an efficient manner.

Results

AI was added to the Autoclave Loading Tool seamlessly. Our planning and research paid off. No rework was needed, usability increased, and time to develop was shorter than if research and design had not been included prior to the build. We have heard reports that it is reducing the amount of time needed to select which parts should run together. No additional training has been requested, indicating a high rate of usability. We are collecting large amount of data on this application and will use it to compare efficiency and error rates. The product has provided business value to the company, reduced time to complete tasks, and has not placed any additional mental burden on users.