PredictiveHire and the AI Recruiting Startup Trend
Table of contents
Ever notice that you can go all your life without ever seeing or hearing about a certain thing but that once you come across that particular thing, you suddenly see it everywhere? Like the actor Ryan Gosling. We never heard of this guy until recently, but now apparently Hollywood can’t make a film without including this dude in the cast. We’ve had a similar experience with artificial intelligence in HR recruiting since first broaching the topic in February. We expressed what could be described as a healthy amount of skepticism that machine learning would have a legitimate role in the recruitment process any time soon, despite what could also be described as a healthy amount of disdain for the whole HR apparatus.
Then, last month, we wrote about seven AI startups that professed to be disrupting the recruiting process. It could as well have been 17 or even 70. Suddenly, everyone seems to be deploying algorithms to find the best and brightest candidates. These AI recruiting startups claim their products make human recruiters more efficient while helping predict who will succeed and who will fail at a particular company. So exactly what kind of AI juju is powering these crystal balls?
Paul Burley, co-founder and CEO of PredictiveHire out of Melbourne, Australia, got in touch with us after reading our June article. The four-year-old startup recently completed a Series A, and has raised $7 million AUD to date ($5.3 million in U.S. greenbacks). It has emerged from “stealth mode,” according to Burley. He was willing to discuss how the AI platform at PredictiveHire can spot the winners from the losers, and why artificial intelligence is suddenly the hottest thing in HR since the performance review.
Under the Hood of PredictiveHire AI
“Under the hood of our platform we use a combination of data science, computer science, behavioural science and statistics,” Burley says by email. “We use advanced data modelling on answers to a short data capture questionnaire, and in some cases business KPIs.”
KPI stands for Key Performance Indicator. If you have spent any time chained to a cubicle then you know that companies are relentlessly assessing success by measuring major objectives, like making money. And making more money. PredictiveHire has about a half-dozen KPI models such as Sales and Culture Fit. It can also mix and match or customize a KPI for a company.
Then the AI goes to work, establishing a “correlation between the questionnaire answers and KPI performance,” Burley continues. Applicants are compared and scored against the incumbent employee pool. Scores range from minus-100 to plus-100, with zero being the baseline average of the company’s current staff. “AI is able to find out what features are statistically significant relating to KPIs used, and then simplify the existing recruitment process in terms of time and effort,” Burley says.
The more data PredictiveHire can accumulate for a particular job type, the better it can predict how well a candidate will jell with a particular company. That’s the machine learning bit. Or as Burley puts it: “Our predictive models are capable of leveraging our accumulated in-depth knowledge embedded in our AI models for certain job families; a new customer benefits from the previous modelling efforts. With applicant data and performance metrics, models use sophisticated machine learning (stacked ensemble) to improve with use and data accumulation over time. Minimum accuracy for deployment is 75 percent.”
Does all that really work? RED SAP Solutions, a global recruitment specialist, reportedly believes so. PredictiveHire built a custom KPI model from the company’s data. During a test phase, company officials found that 90 percent of new hires that left within three months of employment had received a negative prediction from PredictiveHire.
For Simply Business, a large online business insurance broker in the UK, PredictiveHire claims its AI predicted high-performing candidates with 85 percent accuracy. London-based ReThink Recruitment reportedly stopped recruiting low-scoring candidates for its clients after PredictiveHire identified those most likely to bolt within a year with 75 percent accuracy. We find it interesting that many of PredictiveHire’s clients are themselves in the recruiting business, which in itself seems to speak volumes about the recruiting business. But not all: One success story recounts how a “multi-location theme park operator” used PredictiveHire to improve shift attendance among what is probably not the most reliable or highly motivated workforce. Shift attendance was significantly higher for the 76% of hires who received a positive prediction. No stats on how many of them were not getting stoned between shifts.
Just Another Tool?
If AI in recruitment works as well as PredictiveHire claims, then we see opportunities not only throughout the HR process but in totally unrelated businesses. Like online dating. Can you imagine how many more people would sign up for OKCupid if they could be guaranteed potential dates (i.e., candidates) would meet certain KPI such as … well, use your imagination. We are a PG-rated website, after all. Divorce rates would plunge.
Burley at PredictiveHire says one thing that won’t take a dive is the employment of recruiters themselves.
“We understand that there’s some nervousness around the extent to which AI can substitute for human decision-making in HR—and we don’t see PredictiveHire as removing the requirement for recruitment specialists. We view it as a tool that can do one aspect of a recruitment task much more efficiently and effectively and gives them time and resources (effort and cost) back. It also has the effect of making the subsequent tools in the recruitment funnel even more effective, as they are being used on a more qualified group of people.”
AI Recruiting Has Momentum
Just as we were sending emails back-and-forth with Burley, news broke of another AI recruitment startup raising serious cash. New York-based Harver raked in $8.1 million in a Series A led by Insight Venture Partners, TechCrunch recently reported, bringing total funding to $11.4 million. Customers include companies like Booking.com, Netflix, Zappos and OpenTable. Its AI platform, TalentPitch, pitches a familiar proposition by now: machines applying predictive analytics to cull the pool of candidates that will fit best in any given company.
Frankly, as TechCrunch describes the application process, TalentPitch sounds like the AI equivalent of the Tough Mudder obstacle race: “Candidates apply for the job using a variety of modules, including: knockout questions that help work out if this is the right person from the start, then a company/culture video tour, real-life judgment tests, personality testing, intelligence testing, language testing, games that assess candidate skills and more. During the process, Harver’s algorithms gather relevant data and calculate a candidate’s propensity for success in the job role, they say.”
We can’t wait to apply. Still, there’s no arguing that companies are embracing AI for recruitment.
“There is a big opportunity in this space because machines are better at making predictions than people. At the very first stages of recruitment, it’s all about screening people in or out to progress further down the recruitment funnel,” says Burley, a 20-year HR veteran.
Our own prediction is that HR professionals have a buzz on for AI right now, so companies peddling machine learning in this space will thrive—at least for a while. Burley himself concedes that not every startup is on the up and up. And you’ve heard us preach about the pitfalls of believing all the AI hype, but it bears repeating: The litmus test for any “AI company” should be real-world quantifiable performance, scalability and continuous improvement over time.
Become a premium member and get access to hundreds of premium articles, reports and additional content.
Nanalyze Premium is your comprehensive guide to investing in disruptive technologies. Read by the top investment banks, management consultancies, VCs, and research houses. Trusted by over 100,000 institutional and retail investors. Covering disruptive technologies for nearly two decades.