Vervoe’s new AI technology tests applicants' on-the-job skills

A new machine learning technology is helping employers select the most skilled job applicants — without ever having to look at a resume.

Vervoe, an Australia-based startup, this week launched an artificial intelligence machine learning technology that ranks job applicants entirely on the results of a test that determines their on-the-job skills then recommends top candidates to HR managers.

“What we set out to do is make the hiring process about who can do the job, rather than who’s good at getting to the interview stage,” says Omer Molad, chief executive and co-founder of Vervoe.

Vervoe launched last year and currently has around 4,000 users on its platform in 75 countries, including staffing company Allegis Global Solutions, global internet and entertainment company Naspers and digital agency AKQA.

Hannah Szabo, operations manager at Flow, a two-year old content marketing agency based in Denver, says Vervoe helped to tighten up their hiring process, and allows them to immediately screen out candidates who aren’t a good fit.

“We can create assessments that filter out candidates who would have flopped on the job,” she says.

Flow has been using Vervoe for eight months and Szabo says it has saved them time and money. Szabo doesn’t have to take time away from her responsibilities to sort through resumes and she won’t have to hire an additional person to oversee hiring. It’s also made the process more organized from the candidate’s perspective, she notes.

“If it’s a very incoherent, inconsistent sporadic hiring process, as a candidate, you lose a lot of respect for the company,” Szabo says.

Employers pay on average $400 to $500 a month for the Vervoe subscription service that allows them to create talent trials, or job-specific tasks that test a candidate’s real world skills. Employers can use Vervoe to create talent trials that tests an applicant’s knowledge.

For example, Molad says, an applicant for a sales job may have to complete a mock cold call, or someone applying to a retail position may have to provide insight into how they would deal with a difficult customer. The hiring manager then manually reviews the test results and grades the applicant’s performance.

Vervoe’s new artificial intelligence grading feature takes this a step further. The system’s algorithm learns as employers manually grade tests. Once Vervoe has gathered enough data on the particular employer’s preferences, it uses that information to grade on its own. Then the system ranks candidates based on a predicted score, or an estimate of what the employer would give the candidate if they were grading them manually, Molad says.

“The benefit to the company is you now have a fully ranked short list and you can decide if you want to review the top five or 10% of that, and makes life much, much easier,” he says.

Szabo says it generally takes her around 15 minutes to create a talent trial, but this depends on the type of trial, and the level of depth of each question. Multiple choice tests, for example, may take less time to create than other more open ended assessments.

Vervoe is not the only technology available that uses skills assessments to test potential applicants. Platforms like Indeed Assessments also allow companies to build modules to test applicant’s skills. HackerRank tests more specialized skills by allowing employers to assess applicants with coding challenges.

One of the goals Molad and his co-founder, David Weinberg, have for Vervoe is to help minimize gender and racial bias in the hiring process. Molad hopes the skills tests will help to minimize conscious and unconscious bias, a crucial topic as employers work to build more diverse and inclusive workforces.

“We’re just saying, ‘Please don’t look at the candidate on paper, just get everybody to do a whole bunch of activities, see their work, and judge based on that’,” Molad says. “Traditional hiring processes make it more difficult for women and underrepresented groups.”

But data is far from perfect, and research has found that machine learning can still result in unconscious bias. Since the algorithm is learning from human users, it is still possible that some bias could be present in the grading process.

For Molad, the objective is to make the hiring process more automated, but also more human. He hopes the technology will help users make the most informed decisions about the best candidates to hire.

“The way we use AI is making the process more human,” he says. “That’s what we stand for.”