Banks Cash In on AI-Enabled Recruiting
Financial institutions are tapping AI and machine learning to zero in on talent for competitive engineering roles.
This program was produced by the Marketing Department of WIRED and Ars in collaboration with CA Technologies.
During his 18-year career working in risk analysis at JP Morgan Chase, Ed Donner was always on the hunt for the right people. For his last project at the bank, Donner needed to hire about 40 Python engineers in just a few months. But recruitment is an imprecise art.
“We used an army of recruiters,” he says. “We used agencies. We used the tech tools that are out there. And I found that either I was just getting this very light flow of one or two candidates a week, or I was inundated with low-quality candidates with the wrong skill set.”
But if state-of-the-art neural networks can chip away at complex problems involving piles of messy data, why not recruitment? The company that ultimately grew from Donner’s recruitment frustrations, Untapt, may be the first to integrate AI tools into the recruitment process specifically for fintech and financial companies.
“It seems that applying that kind of deep learning to problems of recruitment—you have so much data and much of it is unstructured. It seemed like the perfect fit,” says Donner, co-founder and CEO at Untapt.
Uncovering Hidden Talent
In the HR world, that unstructured data takes many forms: titles and dates on a resume, or keywords on a candidate’s LinkedIn profile. The challenge is extracting deeper meaning from those facts. A LinkedIn profile can show that someone worked in DevOps, but whether they merely partnered with the DevOps team or had a junior DevOps role right out of college takes a level of interpretation that a simple text search can’t easily reveal.
“If you look at a search on a traditional job search, you can only search ‘Python’ and have no way of distinguishing these things,” Donner says.
What might read as trivial or granular to a traditional search engine is where Untapt's tech can find hidden gems of importance in the selection process.
“Everything has algorithms, but it doesn't figure out if you're a scripter or object-oriented [engineer]. You miss a lot of people.”
— Len Langsdorf, PeerIQ
Employers specify precisely what type of programmer they’re looking for and the project they’re working on. Employees upload their LinkedIn profiles and resumes, projects they've worked on and specific classes they’ve taken. Untapt then filters the employee information through a deep neural network using algorithms the company spent two years refining to match skill sets to the type of employee a recruiter is looking for.
It worked for Len Langsdorf, founder and CTO of PeerIQ, a fintech company that allows buyers of bundled loans to see precisely what they’re buying. “Everyone has algorithms,” Langsdorf says of recruitment tech generally, “but it doesn’t figure out if you’re a scripter or an object-oriented [engineer]. You miss a lot of people.”
PeerIQ recently brought on a full-stack developer using Untapt, and the process has been so successful that Langsdorf says he makes offers to about 75 percent of the candidates Untapt refers to him.
Donner reports similarly high figures. When the company started out, he says, about 20 percent of candidates its software spotted were invited for interviews—a common percentage for recruiters. Today, he says, it’s generally between 40 and 50 percent.
The Best Candidate Might Be in the Next Cubicle
Companies could also use the technology internally to surface and recommend employees for different roles.
“Companies have amazing jobs for employees inside the firm, but employees aren’t aware of them,” Donner says, “but they can use Untapt to manage employees within the company.” A few banks are considering a pilot with Untapt to do just that, churning internal repositories of data—from employee skills and qualifications to performance evaluations—to match internal workers with company openings.
If Langsdorf’s success is any guide, those companies might also be able to rely on fewer recruiters to find, vet and hire talent for critical but notoriously difficult-to-fill security and engineering programs.