Critical Thinking About HRM

On another note

AI in talent acquisition: a review of AI-applications used in recruitment and selection

Edward Tristram Albert

Abstract

Purpose – The purpose of this study is to explore the current use of artificial intelligence (AI) in the recruitment and selection of candidates. More specifically, this research investigates the level, rate and

potential adoption areas for AI-tools across the hiring process.

Design/methodology/approach – To fulfill that purpose, a two-step approach was adopted. First, the literature was extensively reviewed to identify potential AI-application areas supporting the recruitment and

selection (R&S) process. Second, primary research was carried out in the form of semi-structured thematic

interviews with different types of R&S specialists including HR managers, consultants and academics to

evaluate howmuch of the AI-applications areas identified in the literature review are being used in practice.

Findings – This study presents amultitude of findings. First, it identifies 11 areas across the R&S Process where AI-applications can be applied. However, practitioners currently seem to rely mostly on three:

chatbots, screening software and task automation tools. Second, most companies adopting these AI-

tools tend to be larger, tech-focussed and/or innovative firms. Finally, despite the exponential rate of AI-

adoption, companies have yet to reach an inflection point as they currently show reluctance to invest in

that technology for R&S.

Research limitations/implications – Due to the qualitative and exploratory nature behind the research, this study displays a significant amount of subjectivity, and therefore, lacks generalisability. Despite this

limitation, this study opens the door to many opportunities for academic research, both qualitative and

quantitative.

Originality/value – This paper addresses the huge research gap surrounding AI in R&S, pertaining specifically to the scarcity and poor quality of the current academic literature. Furthermore, this research

provides a comprehensive overview of the state of AI in R&S, which will be helpful for academics and

practitioners looking to rapidly gain a holistic understanding of AI in R&S.

Keywords Big data, Human resources, Artificial intelligence

Paper type Research paper

Introduction

In 2018, the artificial intelligence (AI) industry was valued at a staggering $1.2tn according

to Lovelock et al. (2018) and 61 per cent of businesses were reportedly using AI

somewhere across their organisation (Narrative Science, 2018). No one could have

predicted the meteoric rise of AI-based technologies to such a high level of ubiquity so

rapidly and so soon. However, the justification for such outstanding growth makes a lot of

sense from a business perspective. AI has the potential to significantly increase profitability

by 30 per cent (Purdy and Daugherty, 2017). Even then, these figures are growing at a rate

so alarming that regulators and academics are struggling to keep up.

Edward Tristram Albert is

based at DurhamUniversity

Business School,

Durham, UK.

DOI 10.1108/SHR-04-2019-0024 VOL. 18 NO. 5 2019, pp. 215-221,© Emerald Publishing Limited, ISSN 1475-4398 j STRATEGIC HR REVIEW j PAGE 215

Despite this promising trend, the field of AI in recruitment and selection (R&S) remains

hugely underdeveloped. On the practitioner side, the literature is overly optimistic and

paints a picture that is almost too positive. While on the academic side, the literature

remains close to inexistent and the scarce literature available is dominated by fictional

credibility (Oksanen, 2018).

The purpose of this study is, therefore, to explore the current use of AI in the R&S of

candidates. More specifically, this research investigates the level, rate and potential

adoption areas of AI-tools across the hiring process.

Methodology

To carry out this investigation, a two-step approach was adopted. Firstly, the literature

was extensively reviewed to identify potential AI-application areas supporting the R&S

process. In light of the scarce and poor quality of scholarly articles for this specific

research area, most of the data were sourced from practitioner reports. So, to ensure

reliability and validity, specific vetting factors were applied including strong references,

credible authors (i.e. experience and education), and absence of bias. Naturally, these

reports have their limitations, which is that the organisations behind them have their own

agendas and are notorious for painting a picture that may not objectively reflect the

reality of AI in R&S.

Secondly, primary research was carried out in the form of eight semi-structured thematic

interviews with different types of R&S specialists including HR managers, consultants and

academics to evaluate how much of the AI-applications identified in the literature review are

being used in practice.

Findings

There seems to be a total of 11 areas where AI-tools can be applied to support the R&S

process. For clarity and concision, these AI-applications areas have been dissected into

Table I. Furthermore, to maintain objectivity and breadth of study, each area of AI-

application has been analysed by focusing on the below questions:

� What (What R&S issue is this AI-application addressing?)

� How (What solution does the AI-application provide?)

� Outcomes (What are the benefits of adopting that AI-application?)

� Adoption (What companies and to what extent are AI-applications being adopted?)

� Vendors (What companies are currently selling these AI-applications?)

Although the breakdown below may give the impression that AI is being widely adopted, a

few cautions need to be made. Firstly, it is unclear how many and to what extent these

applications are being adopted in practice because of the bias behind the reports used.

Secondly, these applications may vary in terms of ROI, level of adoption, opportunities for

growth and costs. Thirdly, the magnitude of each application remains unclear as the scarce

information makes it challenging to paint a clear picture. Fourthly, some may be more

popular than others, some may be more mainstream or cheaper, and some are still

underdeveloped but may exhibit higher potential for adoption.

To some extent, the above applications deliver similar outcomes and in many ways, even

overlap. So, to illustrate, Figure 1 below breaks down where and how these R&S AI-

applications are being used across a standard hiring process, and how they overlap.

However, as previously underlined, these AI-applications were extrapolated from

practitioner reports written by organisations keen to market their knowledge and establish

PAGE 216 j STRATEGIC HR REVIEW j VOL. 18 NO. 5 2019

Table I Areas AI tools can be employed to support R&S

Ai tool Problem Solution Outcomes Adoption Vendors

Vacancy

prediction

software

Spontaneous

resignations increase

costs

Software identifies

employees’ behavioural

data and makes a

prediction on likeliness

to leave

Prediction software

gives a head start,

which reduces these

costs

Improved talent attrition

Improved employer

brand

Reduced time to hire

Large companies

(e.g. IBM)

Data-driven firms

(e.g. Facebook)

High candidate

volume (e.g.

Goldman Sachs)

High turnover (e.g.

Call Centres)

Workday talent

insights

Bamboo HR

Job rate

Monster talent

management

Job description

optimisation

Software

Complex jargon,

boring, indirect

discrimination can be

off-putting

Negatively affects

diversity, applicant

volumes and employer

brand

Software provides

recommendations to

optimise job

descriptions and tailor

the language to

different types of

candidates

Improved diversity

Reduces the risk of

indirect discrimination

Higher candidate

engagement

Cisco

American Express

Johnson & Johnson

Nvidia

Expedia

Evernote

Textio

Three sourcing

15Five

Targeted job

advertising

optimisation

Wrong message to the

wrong audience

through the wrong

channels is a waste of

resources

Using AI, ML and data

insights, firms can

target accurate

recommendations to

relevant candidates

Improves candidate

experience

Maximises chances of

candidate engagement

Minimises advertising

spend

Retail sector

Newton

Netflix

YouTube

ClickIQ

PandoLogic

Recruitz

Appcast

Multi-database

candidate

sourcing

Untapped potential of

suitable passive

candidates and former

employees reduces

talent pool quality

AI-tool scans through

multiple databases

(e.g. LinkedIn,

Glassdoor, indeed,

social media profiles)

much faster and more

accurately than a

human recruiter

Accelerates candidate

sourcing rate

Frees up recruiter’s time

to focus on more

essential tasks

Improves quality and

quantity of talent pool

Intel

eBay

Hilton

Verizon

IBM

Accenture

Warner Bros

Hiretual Pro

Ideal

CV Screening

Software

Reviewing CVs is time-

consuming and costly

Human error increases

as the number of CVs

increases

Software instantly

reviews a large volume

of CVs to filter out and

rank the best ones

Reduces bias and issues

associated with human

fatigue

Improves diversity

Reduces costs

Allows recruiters to focus

on more essential tasks

IBM

LinkedIn

Hilton

Goldman Sachs

Amazon

IBM Kenexa

Ideal.

CVViZ

Zoho Recruit

Talent Recruit

Talent Cube

AI-Powered

psychometric

testing

Outdated, boring and

unengaging tests leads

to negative candidate

experience and

negatively affects

employer brand

Tests use AI to provide

engaging tests

designed to improve

candidate experience

while simultaneously

assessing candidates

Allows recruiters to focus

on more essential tasks

Improves diversity in the

work places

Improves the candidate

to hire (C2H) ratio

Unilever

PwC

Accenture

LinkedIn

Tesla

Arctic Shores

Pymetrics

Knack

Video screening

software

Pre-screening

interviews are costly,

biased and time-

consuming

Software analyses

video interviews to

assess person-

organisation and

person-job fit

Reduces bias and

discrimination

Allows recruiters to focus

on other essential tasks

Improves candidate

experience

Vodafone

Intel

Urban Outfitters

IBM

Hilton

Unilever

HireVue

Montage

Wepow

InterviewStream

AI-Powered

background

checking

Background checking

is time-consuming and

ripe with human error

Leads to problematic

employee termination

downstream

AI software scans

through multiple

databases to verify

candidate details such

as criminal record,

credit rating and

references

Allows recruiters to focus

on more essential tasks

Reduces costs

associated with human

errors

Fortune 500 firms

Financial Firms

Uber

Axa Insurance

BT

McAfee

Check’s

Intelligo

GoodHire

HireRight

Sterling Talent

Onfido

(continued)

VOL. 18 NO. 5 2019 j STRATEGIC HR REVIEW j PAGE 217

themselves as go-to sources. In addition to that obvious bias, another issue is the lack of

reporting accuracy. Indeed, most of them do not give an indication of the extent to which

each AI-application is being used in R&S, and those that do, provide conflicting accounts,

which is understandable considering the challenges that comes with quantifying relevant

measurements indicators. Finally, information expires quickly due to the rapidly changing

landscape, so a figure one day may be different from the next. The list is long, but all this

means, is that these 11 potential applications, as attractive as they may be, may not be

used as much as these reports claim, if at all.

This limitation also seems to be mirrored by interviews carried out with different types of

recruitment professionals. According to respondents, companies are only scraping the

surface when it comes to using AI-applications in their R&S process. By all accounts, the

overall consensus is that there aremore companies using AI in R&S than most people think,

but not as much as people think.

Furthermore, based on interview findings, AI-adopters tend to be large organisations with

an abundance of resources or tech firms with the flexibility to leverage internal talent to

implement these tools or innovative firms with enough ambition to take on first-mover risks.

Even small and medium sized enterprises, which, in the UK for instance, constitute 99 per

cent of all businesses and 60 per cent of the workforce is starting to use AI in their R&S

because of firms experimenting with AI-products tailored to their market niche (ICAEW,

2014).

Despite the relatively high percentage of companies adopting AI in R&S, companies are far

from using all 11 application. Those that are in fact adopting AI in R&S, tend to gravitate

around the same three applications, which by order of popularity are chatbots/CRM apps,

admin-related task automation and screening software (CVs and videos). In addition to only

3/11 apps being adopted, the actual commitment made by those AI-adopting firms remains

Table I

Ai tool Problem Solution Outcomes Adoption Vendors

Employer

branding

monitoring

Reputation affects the

way candidates

perceive a potential

employer

Bad reputation leads to

lower talent pool quality

Software scans through

public data to assess

overall sentiment and

identify weak points in

the hiring process

Stronger employer brand

improves talent pool

quality

Positive image for clients

Reduces T2H, staff

turnover and overall

costs

McKinsey & Co

Oracle

HP

Dominos

Lexalytics

Semantria

Microsoft

Thematic

DiscoverText

Candidate

engagement

chatbot/CRM

Direct recruiting and

relationship

management are costly

and time-consuming

Unpredictable or high

volume can lead to

longer responses,

dissatisfied candidates,

which negatively

impacts employer

brand

Chatbots are tool that

leverages Natural

Language Processing

to mimic human

conversational abilities

and can be used to

engage candidates,

provide quick

responses to questions

anytime

Reduces T2H

Allows recruiters to focus

on more essential tasks

Improves candidate

experience and

employer brand

Sephora

eBay

H&M

Pizza Hut

Burberry

IBM

Nuance

NextIT

Kore

Inbenta

Personetics

Aivi

Mya

Beamery

Automated

scheduling

Scheduling calls, tests,

interviews or meetings

is time-consuming and

non-essential

AI system that picks up

on scheduling

expressions to

automatically execute

these admin tasks

Allows recruiters to focus

on more essential tasks

AT&T

Disney

Coca-Cola

Walmart

General Electric

Survey Monkey

X.ai

Troops

Tact

Olono

PAGE 218 j STRATEGIC HR REVIEW j VOL. 18 NO. 5 2019

extremely low. Most companies are piloting these applications rather than actively using

them.

Considering the alleged benefits of implementation, such a low level of adoption is

disappointing. However, the reason behind it makes sense. These AI-tools have only

started to build up momentum over the past few years. As a result of this, most AI-

applications are still at an embryonic stage, present a multitude of technical and human

challenges, and need further development.

Recommendations

Based on these findings, this explorative research posits the following recommendations for

practitioners.

1. HRManagers: How to navigate this constantly changing the landscape?

Figure 1 Use of AI applications in R&S

JJob Offer Extension

Employee Onboarding

R E

C R

U IT

M E

N T

S E

L E

C T

IO N

Vacancy Anticipation

Job Description Formulation

Vacancy Advertising

Candidate CV Screening

Candidate Psychometric Testing

Interview (Video/F2F/Phone)

Assessment Centre

Final Interview (Video/F2F/Phone)

Background Check

RECRUITMENT & SELECTION PROCESS APPLIED AI TOOLS

Targeted Job Advertising

Candidate Sourcing

Pre-screening of CVs

Psychometric Testing

AI Powered Video Interview

AI Powered Background Check

AI TOOLS INDEX BIG DATA AI CAPABILITIES Employer Branding Monitoring Past Data

External Data

Current Data

Machine Learning

Deep Learning

Neural Networks

Data Mining

NLP

Candidate Engagement Chatbot

Automated Scheduling

Vacancy Prediction

Job Description Optimisation

VOL. 18 NO. 5 2019 j STRATEGIC HR REVIEW j PAGE 219

� Recommendation one: be cautious of buying AI-products from vendors as they

may present some technical issues. Ensure meticulous vetting of products before

purchase.

� Recommendation two: precaution is good but late adoption is bad. Preserve a

front-row seat and eyes on the ball to optimise the timing of entry into AI for R&S.

Indeed, there are first-mover disadvantages, but this resource (human talent) is

harder to catch up with.

� Recommendation three: sponsor data scientist training courses. It will elevate your

profile in the industry, you will develop contacts and have first pick on the talent

needed to build in-house solutions.

2. Vendors: How to survive in this changing environment?

� Recommendation one: clients will always be sceptical of your products. To

address that, convince the people these companies listen to, namely, big

consultancies like the Big 4 who advise most Fortune 500 companies.

� Recommendation two: to stay relevant, make it challenging for companies to

replicate your product in-house. The following strategies might help. Firstly, make

you are offering so compelling (value for money) that it defeats the point of

replicating it in-house. Secondly, forge and maintain strong relationships with key

industry players (i.e. competitors, suppliers, governing bodies, etc). Thirdly,

establish yourself as a go-to player. Finally, ensure your workforce is happy

enough to reduce the risk of intellectual property poaching.

3. Recruitment Agencies: your existence is threatened by AI-vendors with better offerings

and/or companies building cheaper in-house solutions. Conveniently, some companies

lack the confidence to undertake that transition, and there will still be a need for human

interaction in R&S, especially around senior vacancies. In the interim, below are some

suggestions:

� Recommendation one: think about integrating AI into your business model.

� Recommendation two: stay a step ahead of AI-vendors by offering a more

compelling product.

� Recommendation three: hire a tech-savvy workforce to reduce resistance to

change and stimulate innovation.

4. Candidates: AI in R&S is coming whether you like it, and revolutionises the way you

need to present yourself, which is good news for minorities and disadvantaged groups

(Min, 2017):

� Recommendation one: research how to please the machine and the interviewers.

Find out what the system looks for to increase your chances of getting hired (e.g.

displaying appropriate body language in video screening software).

“Considering the alleged benefits of implementation, such a low level of adoption is disappointing. But the reason behind it makes sense. These AI-tools have only started to build up momentum over the past few years”.

PAGE 220 j STRATEGIC HR REVIEW j VOL. 18 NO. 5 2019

5. Entrepreneurs: AI in R&S is filled with unexplored niches, to name of few:

� Industries (e.g. call centres, tech and sport).

� Company sizes (e.g. large vs SME).

� Labour skill type (e.g. coders, yacht-crew, waiter).

� Geographies (e.g. US/UK/China vs unexplored markets).

Furthermore, companies are acquiring vendors rather than building in-house tools, which

presents a buyout opportunity. However, an eye needs to be kept on important aspects

such as changing regulations, ethical dilemmas and threat of new entrants.

References

ICAEW (2014), “The 99.9%: small andmedium-sized businesses: who are they and what do they need?”,

available at: www.icaew.com/-/media/corporate/files/technical/business-and-financial-management/

smes/sbm/the-99-percent-small-and-medium-sized-businesses-finalpdf.ashx

Lovelock, J.D. Tan, S. Woodward, A. and Priestley, A. (2018), “Forecast: the business value of artificial

intelligence, Worldwide, 2017-2025”, Gartner, available at: http://k1.caict.ac.cn/yjts/qqzkgz/zksl/201805/

P020180504572266109739.pdf

Min, J.A. (2017), “4 Promising ways AI is helping diversity in recruitment”, Ideal, available at: https://ideal.

com/ai-diversity-recruitment/

Narrative Science (2018), “Artificial intelligence (AI) adoption grew over 60% in the last year”,

GlobeNewswire News Room, available at: www.globenewswire.com/news-release/2018/01/17/1295827/

0/en/Artificial-Intelligence-AI-Adoption-Grew-Over-60-in-the-Last-Year.html

Oksanen, R. (2018), New Technology-based Recruitment Methods, University of Tampere. available at:

https://tampub.uta.fi/bitstream/handle/10024/103591/1527751872.pdf?sequence=1&isAllowed=y

Purdy, M. andDaugherty, P. (2017), “HowAI boosts industry profits and innovation”, Accenture, available

at: www.accenture.com/fr-fr/_acnmedia/36DC7F76EAB444CAB6A7F44017CC3997.pdf

Corresponding author

Edward Tristram Albert can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website: www.emeraldgrouppublishing.com/licensing/reprints.htm Or contact us for further details: [email protected]

VOL. 18 NO. 5 2019 j STRATEGIC HR REVIEW j PAGE 221