- Internal Administrators
What I did
- User & Product Research
- UI / UX Design
- User Testing
Questions and Answers, all in one location.
The Duke Energy Smart Search Assistant (DESSA) Dashboard is an tool built to support the Smart Search Engine. The Smart Search Engine incorporates machine learning to more accurately predict and assist customers as they search the Duke Energy site.
DESSA allows the administrator to review performance, data and make changes to questions and responses that are served up.
Defining a goal
Why did we do this?
There were multiple reasons that Duke Energy chose to build DESSA
- Replace the vendor search engine costing Duke Energy $200,000 a year
- Create a search engine replacement w/ added capabilities
- Create a platform to control DESSA
The scope of this project is to provide a dashboard that allows a user to review DESSA’s performance and make changes to the database that can enhance it’s response.
Constraints were low with the project receiving excellent funding and having great support from the business and stakeholders.
The biggest constraint from a UX perspective was time. I had to gather requirements from the Product Owner, run design thinking sessions and at the same time learn about machine learning.
All this had to be accomplished in a short amount of time as the team was already working on setting up backend services.
In order to ensure all requirements were documented, I held whiteboard sessions with stakeholders, SME’s and product owners.
The most important concept for DESSA is to give users a simple way to access and update constantly changing information. I gathered ideas from other products such as Gitlab that took data and laid it out into an easy to understand manner.
To keep things simple, I prefer to use an app called Mindnode for information architecture. It’s simple to use and easy to decipher.
Whiteboarding sessions become wireframes.
Go for launch
After a 72 hour transition off the old architecture, DESSA was launched in the spring of 2019 and the dashboard was up and running.
Lessons and Outcomes
This project was one I quite enjoyed. The team was amazing and I learned a ton about chatbots and how AI works.
The technology worked flawlessly and our internal media reported it as one of the smoothest launches in recent memory.
Still there were some areas that could be improved:
First, the information architecture could use some structural changes. I relied to heavily on tabs when I could have combined them into one section.
Automating the pkl file drop would help prevent errors with a user uploading the incorrect file. Granted, this was a limitation of the project but for future releases this is something that could have been addressed.
Overall, the project had one of the most successful launches in the Idea Lab’s history. Not only did it save Duke Energy $200,000 per year in replacing the aging vendor architecture, the underlying technology will set Duke up for future success with better serving their customers through predictive search capabilities, rich data analytics and machine learning tech.