AI Workbench Concept
WINTER 2019 - 2020
How might a minimally-technical corporate workers use AI to solve problems and find answers in their job?
Product Definition, Prototyping, Proof of Concept
One of Augustus' main goals is to make AI more accessible to a greater variety of members of the global workforce rather than just developers and people in highly-technical roles. The vision was to make an AI product that is as easy to use as Excel, where a user could string together AI models to solve problems best suited for AI. There was ambiguity and confusion at the company on how this product would function, look, and feel, so in order to create a clearer product and workflow definition, I challenged myself to think of a use case and for how someone could use this AI product to answer an everyday corporate question.
USE CASE DEFINITION
Since Augustus' focus has been computer vision applications, I wanted to come up with a use case for someone working in a visual media industry may have. I chose the advertising industry since high volumes visual brand exposure is key to their company's success. I turned this concept into the following persona + use case:
Sabrina is an Ad-Effectiveness Analyst on the Digital Marketing team at Coca-Cola.
GOAL & TASK
Her team recently launched a marketing campaign for Coca-Cola glass bottles via Instagram and her manager wants to know if it increased organic brand sharing of the product on social media. Sabrina is tasked to determine the prevalence of Coca-Cola bottles in Instagram photos since the launch of the campaign versus that of their competitors.
PLAN OF ACTION
Sabrina knows that while she can use a program to look for #cocacola tags and scan Instagram captions, she has no way of feasibly checking a large volume of photos for the Coca-Cola bottles. She realizes AI is a good fit for scanning photos and makes the following plan:
Use an internal tool to scrape Instagram for thousands of photos.
Set up and test a workflow in the Augustus AI Workbench that finds Coca-Cola and brand competitor bottles in photos and exports the results to a spreadsheet for later analysis.
Run this AI workflow on the bulk amount of Instagram image data that she acquired in the first step to create the results spreadsheet.
Work with the spreadsheet data to find the number of unique photos with Coca-Cola bottles and competing bottle brands.
I then made a demo for how Sabrina may use our product during step two of the process.
USE CASE DEMO
I presented these mockups and demos to multiple product managers and the CTO of the company. I received positive feedback on the user flow and visual design, as well as confirmation that these designs reflected the direction the company should go. However, the company needed to redirect resources to developing its existing products, and tabled the concept for now.
If I had more time and resources, I would have worked with the product team to identify 1-2 target industries with high-value business problems that could be solved with AI. I would then model solutions to these problems with our AI engineers and brainstorm with them on the core features our product would need for a user to create these solutions. At this stage, I would suggest that we start building and using the product as an internal tool to solve clients' business needs, as well as conduct user research with some of our clients' employees. With this combination of heavy internal use on production applications and external user research, I would be able to rapidly identify and test the usability of fundamental product features, while understanding future users' priorities, workflows, and technical ability to inform later iterations and simplifications of the product.