From Unilever to Goldman Sachs, big companies are turning to AI to hire new recruits. The results have been promising, creating workplace diversity and faster recruitment, but there are still kinks that must be ironed out, writes Ross Davies.
he hiring process can be laborious for applicants and companies alike. While it may have been expedited by the birth of
the internet, jobseekers still face the painstaking tedium of endless form-filling – not to mention getting their CVs and covering letters just right. HR departments must then go through stacks of applications to separate ‘the wheat from the chaff’.
That is even before the interview stages – a procedure that can take months when a high-level position needs to be decided upon. This process uses precious time that neither party can afford to waste.
It was only a matter of time before businesses began looking at AI as a means of condensing their recruitment measures into something less time-consuming and more streamlined.
One such group that has done so is Unilever, which began using AI to hire entry-level employees in its US operations over the past year.
Mike Clementi, vice-president, human resources – North America and global customer development, recently stated that the catalyst for the move was a creeping sentiment in the boardroom that traditional recruitment methods, such as sending reps to sniff out talent at elite universities, had grown stale.
Instead, the consumer goods group – which has 170,000 employees on its books globally – decided to partner two digital HR service providers to replace manual overtures with an AI screening process.
Making a game of it
The process goes as follows: candidates receive notifications of openings through social media and professional networking outlets, and can apply by sending a link to a LinkedIn profile. CVs are not required.
After which, applicants play 12 neuroscience ‘games’ that test focus, memory, tendency for risk and emotional intelligence. This stage takes less than 20 minutes.
If successful, candidates then move onto a preliminary interview, using a tool, through which they respond to preset questions on a smartphone or tablet. The technology then provides an analysis – based on keywords, intonation and body language – to be used by Unilever’s HR department.
If the applicant makes it through the two steps, they are invited for an immersion day, in which a hiring manager gives a decision at the end of the session.
According to Unilever, the new strategy has been a quantifiable success. Results shared by the group earlier last year revealed that between June 2016 and July 2017, job applications doubled within the first 90 days from 15,000 to 30,000, year on year.
The average time span for hiring dropped from four months to four weeks, saving around 50,000 hours of applicants’ time. For recruiters, time spent going through applications was slashed by 75%.
AI recruitment has also brought about greater diversity in Unilever’s workplace, with a 209.5% rise in the employment of non-white applicants.
Unilever is not the only big hitter that has sought to digitise its recruitment. On Wall Street, Morgan Stanley, Citigroup and Goldman Sachs have all implemented AI software. There is certainly plenty of evidence to suggest that an appetite is building for algorithms to replace CVs.
According to a survey by Ideal, a Toronto-based builder of recruitment automation software, more than half of talent acquisition leaders claim the hardest part of hiring is the screening process due to large pools of applicants; in fact, 75–88% of CVs received were from unsuitable candidates.
This is compounded by the fact that in some sectors – particularly tech – hiring volumes are creeping up.
More with less
A recent LinkedIn survey on global recruiting trends revealed that 56% of managers expected their hiring volumes to increase this year, despite two thirds claiming that their recruitment teams would either stay the same, or contract. In other words, recruiters are in a quandary of having to do more with less.
While AI can help in reducing time-to-hire periods – meaning companies do not lose out on talent to competitors that are faster on their feet – it can also improve the quality of ‘new blood’.
"Quality of hire used to be a bit of a recruiting KPI (key performance indicator) ‘black box’, due to an inability to close the data loop – such as measuring what happens to the candidates after they get hired," reads a definitive guide to AI recruitment on Ideal’s website. "As HR data has become easier to collect, access and analyse over the years, quality of hire has become recruiting’s top KPI. The promise of AI for improving quality of hire lies in its ability to use data to standardise the matching between candidates’ experience, knowledge and skills, and the requirements of the job."
According to the Ideal survey, results in the field have been promising. Early adopters of AI-powered software have witnessed the cost of each screening drop by 75%, with staff turnover falling by 35%. Revenue per employee is also up 4%.
However, there are challenges to overcome: firstly, AI requires a lot of data to be able to accurately mimic human intelligence. In the case of the screening process, software may need to have processed hundreds of thousands of CVs just to create a baseline from which to judge a candidate’s suitability.
Despite the improvements in diversity reported by Unilever, some are concerned that AI – and its ability to assimilate human biases – could be used to potentially reinforce existing inequalities in some workforces.
If a company has long standing bias against age, gender and race, then it will be reflected in its recruitment history – the very history from which AI is required to learn from.
"We know that AI is not neutral," Y-Vonne Hutchinson, founder of the start-up ReadySet, told the Financial Times in September last year.
"AI is designed by people and those people have their own biases, and sometimes those biases get embedded into the tools and the platforms they use," she added.
Hutchinson’s point is salient. Last year, AI predictive software used by some US courts for parole decisions, was reported to discriminate against African-Americans. The investigation, led by journalism body ProPublica, found the technology incorrectly deemed black prisoners twice as likely as white defendants to reoffend.
"To avoid replicating any biases that may already exist, make sure the AI software vendor you use is aware of these issues and has taken steps to remove clear patterns of potential bias," advises Ideal.
Bots to come
Innovations in AI software are continuing apace. Recruiters have also started to look to chatbots – which are already a common feature in retail and marketing – to provide real-time interaction with candidates throughout the application process.
As well as helping hirers to get a better sense of who they might choose, chatbots could also improve the candidate’s perception of a company.
According to a 2015 study by an employment website, 58% of jobseekers reported having negative perceptions towards companies that failed to respond directly to applications, whereas 67% favoured firms that provided regular updates throughout the application process.
AI remains a divisive issue, with Professor Stephen Hawking warning in 2014 that it could spell the end of the human race. Jobseekers are certainly likely to have misgivings about their professional fate being at the mercy of an algorithm. Within HR circles, it will also not be to everyone’s tastes, states Ideal in its guidelines. "HR professionals are often bombarded with the latest and greatest trend that disappears just as quickly," it reads.
"Understandably, recruiting and talent acquisition leaders can be sceptical of any technology that promises to make their jobs easier. They want to be sure that any software that will automate one of their work tasks is going to be able to do as good of a job as they can."
Perhaps hiring managers may be more favourably disposed to another term: augmented intelligence. This concept supports technology that compliments human capabilities, as opposed to replacing them.
Some areas of research support this theory. According to a study by McKinsey & Company, while 49% of the activities workers undertake today could be automated with AI, fewer than 5% of jobs could be fully automated.
Using the internet for applications has already driven a ‘coach and horses’ through the paper-based hiring process of yesteryear. AI, therefore, has the potential to move things even further along for employers.