Filling large numbers of front-line customer-facing positions, in areas such as call centers, is a time consuming and often inaccurate process. But an Australian company has developed a text-based artificial intelligence test which presents a powerful potential solution.
PredictiveHire is a Melbourne-based company that offers a text-based recruitment tool. From five questions posed in a chat session, the AI can extract 80 different features about the application, from the applicant’s personality through to their English proficiency through to how likely the person is to stay in the job.
Barb Hyman, PredictiveHire’s chief executive, said the company’s approach not only saves time and expense but eliminates bias and delivers a better result in terms of hiring yield.
“We have invented a whole new way of understanding people through text, through responses to five structured questions delivered over a mobile phone,” she said. “If we give you ten people as a hiring manager, you want to hire eight of them, and that is a great result.”
PredictiveHire, in which global recruitment firm Hudson recently took a stake, started the development of its AI with a database of 11,000 words around 18 months ago, but that has now grown to 60 million words. In another 18 months, the aim is to build it up to 1 billion.
In comparison, Hyman pointed to Wikipedia, which she says has 16 billion words in its database after 18 years of operation.
As the database expands, so does the accuracy of the AI and its ability to find the right candidates. For clients, the yield has improved from around 65% of applicants to about 85%. At one client, a major airline, hiring yields improved from approximately 30% to 75%.
To improve accuracy, PredictiveHire then looks at the correlation between who gets hired and who then performs well in the role, with this information going back into understanding candidate responses.
“What is really disruptive about AI in this space is that it is based on a performance profile rather than a hypothesis on whether or not a person is going to be good in a job,” said Hyman.
“Over time, it builds the ability to reduce churn, and it really functions as a bridge between the business and HR.”
The solution doesn’t merely find candidates for jobs on behalf of clients. The AI gives feedback to the candidates, regardless of whether they have been successful or not.
“This is pretty unique,” said Hyman. “Every candidate receives automated personalized feedback with some coaching.”
“This is completely transformational, and is there to help people build their self-awareness and confidence, and the capacity to sell themselves at an interview,” she added.
Diverse recruitment uses
Most corporate clients use PredictiveHire instead of the CV screening process and use it as a way to shorten the field of interview candidates.
Some clients are also adapting the solution for internal use to stay in touch with staff during the COVID-19 pandemic.
“It can be a useful tool to stay connected with the furloughed workforce,” said Hyman.
“The employee gets this ‘sugar hit’ of awareness as the employer reaches out to them, but the responses also give the employer a good idea of how the employee is feeling and coping.”
Hyman said PredictiveHire has ambitions to develop its product to a point where it can be a “personal AI coach.”
“Our vision is that ultimately people can have this personalized AI coach to help them with their soft skills,” she explained.
“That is where we want to get to, and I think it is massively democratizing. We are like ‘Moneyball’ for people. It’s how do we look beyond the obvious, and really get to know people and part of what we are discovering is that many people don’t know themselves.”
The AI, she said, can identify when people are going for the wrong job, and help them “self-discover” what they are best suited to.
Explainable recruitment AI
Another is transparency. PredictiveHire does not use deep learning or factor in other data from sources such as social media.
“We only use features we can explain which are visible to the client,” said Hyman. “The only data we use is what people themselves write, and that creates trust and confidence.”
Photo credit: iStockphoto/Andrii Yalanskyi