A city looking to attract talent to the area offers a one-year remote work program that offers cash and co-working space in exchange for relocation. The program receives 1,000s of applicants in response to their marketing efforts.
A city's remote work program that offers cash and co-working space in exchange for relocation often receives 1000s of applications at a time. The application is over 200 questions long and consists of many open-ended questions as they are trying to find a very specific type of applicant. They look to select a diverse group of people most likely to follow through with moving to their city, working in the city, and even possibly staying long term.
The popularity of the program means that the review team gets inundated with applicants every time they publish a marketing campaign. It was often taking them weeks to catchup on reviewing all of the applicants they receive. They make sure to dedicate time and attention to each application so that everyone gets a fair chance. Filters help to narrow the scope of review somewhat, but they were still left with a lot to do. With existing tools, their ability to meaningfully sort and prioritize the applicants that fit their criteria was limited.
This manual process was time consuming and inefficient. A new solution was needed to help make sure very applicant got reviewed in a shorter amount of time.
The solution leveraged the data they had already been collecting and creating with their manual process. The applicant strength ratings that reviewers assigned became the labels Vettd would use to train a custom AI model. Vettd used the previous year’s applicant scoring to create an AI model that accurately replicated those scoring assignments.
The data from their manual review process was incredibly complete and thorough. After assessing their data situation, Vettd recommended simplifying the AI’s interpretation of the strength rating. This would lead to reaching the highest possible accuracy rates without sacrificing the usability of the model.
Our customer ended up with an AI model that identified strong and weak applicants with near 100% accuracy. We worked with their development team to deploy the model and push results into Airtable, their applicant review tool. As new prospects submitted applications, Vettd’s model would analyze the natural language and populate model results directly in their existing workflow.
Within 5 weeks, our customer had AI evaluating every incoming applicant ensuring that everyone is thoroughly considered. The model helped them identify which applicants to review first, saving them a lot of time in their manual review phase.