Talent Acquisition Tech: It’s About Automation, Not AI. Yet.
At HR Tech 2018 “Artificial Intelligence” was the phrase of the week, slapped onto as many tag lines as possible. But after demo-ing many technologies at the Expo, I’m not convinced we’re there yet. HR tech tends to lag most other industries, and according to research by Future Workplace only 6% of TA leaders are using AI. Instead, there is still plenty of work to do around automating simple and repetitive aspects of the recruitment process, which make up approximately 70% of a recruiter’s job. The majority of process automation technology today sits in three categories: chatbot, scheduling, and advanced screening.
Chatbot companies, such as Mya, Wade and Wendy, Paradox, Ally-O, TextRecruit and Canvas, exist to relieve recruiters from spending time on mundane tasks but more importantly, the goal is to provide a better candidate experience by surfacing positions relevant to their skills and interest, shortening the response time, and interact with candidates outside of business hours. These bots range in capability, from canned Q&A to more sophisticated NLP and responses. While the concept of a human-like bot having meaningful conversations with candidates is enticing, the reality today is well short. Many of these companies still rely on humans to take over conversations the bots can’t understand. Instead, the play is to “save” all candidate interactions as data points that can be later reviewed by a recruiter.
Many TA leaders I spoke to understand this reality however, and instead find value in the ability to save a recruiter’s time on the front end while increasing their top of funnel activity and quality by simply having a tool to capture candidates at the career landing page stage. Addeco is already seeing results in this regard: 89% of candidates (vs. 43% prior to chatbot) filled out a prescreen and the applicant to hire ratio reduced from 10:1 to 7:1.
Scheduling is a complex use case to solve, particularly at scale, but can save significant recruiter time or reduce a company’s need for recruitment schedulers (yes – that’s a job position!). Companies such as Timetrade, Calendly and x.ai have been working on meeting scheduling beyond the HR use case and it’s not easy. X.ai employs hundreds of humans in the Philippines to continue training its AI on “edge cases” such as: time zones, recurring availability, generic references to dates, places and names, and meeting formats. Timetrade has focused its product development on serving complex and large scale use cases, where the challenge is backend matching of skill, location, nuanced availability and other requirements. Thankfully, the recruiter scheduling use case is narrower and therefore has fewer edge cases to account for. Companies such as GoodTime and IntervewierAssistant focus solely on the recruiting use case, but are faced with the additional complexity of matching candidates with interviewers based on availability, skill set and progression in the interview process. I met several TA leaders who, when faced with the build vs. buy decision, ultimately decide building is too much of a strain on IT resources given the complexity of the technology.
The last automation category is perhaps the most advanced in comparison when it comes to artificial intelligence development: advanced screening. This includes video assessment companies like HireVue, in which an AI analyzes answers, tone, and body language. Other advanced screening companies such as Pymetrics, HackerRank and Hire-Intelligence have developed proprietary tests to surface the best fit candidates for any given role. In a Business Insider article, Unilever North America shared preliminary results from working with advanced screening tools. In a year, they experienced double the number of applicants 90 days after posting a job, reduced time to hire from four months to four weeks, increased offer acceptance rate, and hired its most diverse class to date.
AI Resulting in Diversity?
Speaking of diversity, there is a lot of discussion today on whether artificial intelligence can eliminate bias from hiring. Unless programmed otherwise, automation could filter out data such as age, gender and race and instead focus on experience, competency, skills and performance. However, there is an interest debate in the AI community about which algorithms have the potential to make hiring more fair. Some models were built with inherit bias in them, since they are built by programmers who have prioritized certain qualities over others. Technology alone isn’t enough to solve the diversity issue, but automating pieces of the recruitment process may yield candidate pools more representative of the general population.
Automation to AI
Where is HR tech headed? We are just starting to see early applications of predictive and prescriptive analytics, powered by machine learning. The goal is for a machine to be able to surface the best candidate for a role or prescribe the best recruiting process to achieve lower applicant to close ratios. I haven’t yet met a company with enough data or proof points to show consistent results, but this type of technology has a clear ROI.
It’s important to note that TA leaders strongly agreed with the notion that despite the positive early results from automating the recruiting process and the promise of real AI on the horizon, there is no substitute for the human interaction. Where human interaction is necessary versus better replaced by machine is a matter of philosophy, but the HR department is not immune to this digital transformation experienced by all other industries. It is the startup’s job to continue pushing the boundaries of intelligent systems. Until there more progress on AI for HR tech, TA leaders are testing automation solutions to solve problems of today.