Rachana Rao and Becky Hill 29th October 2019 Executive Summary Question How is the role of AI in talent acquisition evolving and at which points/aspects of the TA lifecycle is AI being used (what kind of products and vendors exist)? Overview Artificial Intelligence (AI) is a rapidly emerging technology that is at the forefront of industry innovation and efficiency. Despite the rapid technological advances and energy surrounding AI’s potential, there are pitfalls in terms of AI’s application in the Talent Acquisition (TA) process and many companies have yet to identify strategic points of AI implementation. Through our research, we have identified a variety of AI products available at specific points in the TA lifecycle. We weigh the most apparent drawbacks of using AI in TA as it currently exists, and share research and benchmarks of companies implementing AI in TA. We have identified key areas of AI use in the TA lifecycle, and evidence as to the utility and validity of the technology use along with predictions for future engagement with AI and preparing your workforce. Present AI Available and in Use: The use of AI in talent acquisition is currently low, as reflected by only 10% of respondents to a broad Oracle survey to HR practitioners who indicated they made high use of tech in recruitment. However, organizations expect to drastically increase their utilization of AI in the next two years3. We found many resources on the various technologies being utilized in TA2, 9, 12, but there is not one comprehensive resource detailing every vendor and tool available at specific stages in the TA lifecycle. We attribute this to the rapidly evolving nature of AI. We discovered impactful resources including: ● A definition of the 11 areas across which AI can be applied for recruitment and selection, including companies currently employing the tech and vendors offering it5 (Appendix A). ● An illustration of the application of AI tools for TA, which guides the consumer as they engage with the tools5 (Appendix B). ● Various vendors identified with AI tools available to be applied to specific stages of TA1, 6 (Appendices C and D). Challenges with Current AI Technology: One ongoing challenge since the application of AI in TA has been the programming of technology to replicate human decision-making, which can result in the perpetuation of implicit bias11. When algorithms or machine-learning tools are programmed poorly there is potential for them to perform accordingly, discriminating against candidates based on gender, ethnicity, etc. Additionally, the cost implications and lack of understanding of business benefits and value currently act as barriers for the extensive usage of AI in TA3. Lastly, the growing use of AI in TA requires people who know how to engage with and implement AI. There is currently a scarcity in the HR realm of professionals who are skilled in AI3. In the future, there is projected to be a tremendous AI skills gap in the workforce11. Utility: ● A 2019 broad survey of HR professionals indicate that the most useful AI-related tools according to their ability to improve talent acquisition are: 1. Big-data analytics 2. Automated assessments 3. Predictive analytics3 (Appendix E) ● An empirical examination of the exchange between predictive hiring algorithms, recruitment and the data-driven nature of AI predict the “most useful interviewing innovations... and areas where [AI] will impact recruiting”7 (Appendices F and G). ● Substantial data analysis and estimates on the utilization of AI in TA processes in the interviewing and assessment of candidates (downstream engagement impacts) and gross value added by country provide guidance for strategic AI implementation13 (Appendices H and I). ● It is important to establish the validity of an AI technology by using data and empirical analysis - experimentation, even - to challenge causal assumptions, garner buy in and determine a path forward14. Even if utility and validity is established for a specific AI tool, it is in a company’s best interest to verify their implementation of that technology. Countering Bias: ● Although there is evidence of AI perpetuating bias in the TA process, the algorithmic screening technology is still significantly less biased and more accurate than its human counterparts4. The technology is only as good as the qualifications you provide, however, so the utility of the tool cannot be guaranteed if intentional crafting of unbiased criteria for a position has not occurred. ● In order to combat bias being introduced or unconsciously programmed into the algorithms of AI technology being used for TA, Microsoft is developing AI to remove bias in other AI technologies11. Business Cases and Tools: ● Hilton has been using AI technologies from brands HireVue and AllyO in the initial stages of the recruitment process for call-center reps, which has allowed the company to make: ○ 400% more offers with 23% less staff ○ Decrease time to fill from six weeks to one week ○ Better hires - more top-performers - reducing turnover by using AI predictive insights from video interviews15 ● Companies who are currently piloting AI in recruitment and selection (most are piloting rather than fully adopting) are tending to do so using three key applications, despite the 11 different applications available: 1. Chatbots/CRM apps 2. Admin-related task automation 3. Screening software (CVs and video)5 Upskilling in Data: ● “63% of companies surveyed by Forbes are now providing in-house data analytics training” so that even if employees cannot develop the technologies, they can at least use them effectively8. Since about a third of AI-related positions require a PhD, extensive years of experience, and often come at a very high cost, organizations are urged to hire employees with a bachelor’s degree in mathematics or physics with the intent to further train these individuals in AI8. ● It is important to implement a sophisticated, strategic data strategy in conjunction with AI adoption; know your data points and how to leverage them10. Additionally, it is imperative to ensure you’ve thoroughly vetted any vendor prior to investing in their AI tools, and to ensure you have a large, quality data set5. Conclusion AI has relevant applications to the talent acquisition process and provides streamlining that benefits businesses and candidates. However, limited research has been conducted as to the utility of various AI tools as they apply to TA13, which was reinforced throughout our process. Companies are urged to cautiously vet AI products prior to implementing them in their TA processes, while also not pumping the brakes on AI entirely, as early adopters are projected to have a significant advantage when compared to companies who wait to adopt AI tools5. Using the research that is available to make an informed decision to implement AI tools in the TA lifecycle while carefully monitoring the resulting data is the approach most companies are taking. References 1. (2018) Gartner Inquiry - Redacted for anonymity of CAHRS partner company 2. (2018) Prepare for the jobs and skills of the future with IBM Watson Talent Frameworks, ibm.biz/talentframeworks. 3. (2019) The 2019 State of Artificial Intelligence in Talent Acquisition, Oracle, hr.com. Retrieved October 2, 2019, from https://www.oracle.com/a/ocom/docs/artificial-intelligence-in-talent- acquisition.pdf?elqTrackId=1279a8827f3d4548ae3f966beeeef458&elqaid=83148&elqat=2 4. Ajunwa, Ifeoma, The Paradox of Automation as Anti-Bias Intervention (March 10, 2016). Ifeoma Ajunwa, The Paradox of Automation as Anti-Bias Intervention, 41 Cardozo, L. Rev. __ (2020 Forthcoming). Available at SSRN: https://ssrn.com/abstract=2746078 or http://dx.doi.org/10.2139/ssrn.2746078 5. Albert, E. T. (2019). AI in talent acquisition: a review of AI-applications used in recruitment and selection. Strategic HR Review, 18(5), 215–221. https://doi.org/10.1108/SHR-04-2019-0024 6. Bist, Bhawna & Jackson, Mike (2017). Colgate HR2020 Digital Acceleration (copyright 2017, Deloitte Development LLC.), 10. 7. Bongard, A. (2019). Automating Talent Acquisition: Smart Recruitment, Predictive Hiring Algorithms, and the Data-driven Nature of Artificial Intelligence. Psychosociological Issues in Human Resource Management, 7(1), 36–41. https://doi-org.proxy.library.cornell.edu/10.22381/PIHRM7120193 8. Boyd, Clark (2017). AI scientists: How can companies deal with the shortage of talent? Towards Data Science.https://towardsdatascience.com/ai-scientists-how-can-companies-deal-with-the-shortage-of- talent-11ab48566677 9. Cerrato, J. , Freyermuth, J., Kostoulas, J., Poitevin, H., Hanscome, R. (2018). Cool Vendors in Human Capital Management for Talent Acquisition. Gartner 10. Chui, M., Manyika, J., Miremadi, M. (2018). What AI can and can’t do (yet) for your business. McKinsey Quarterly, mckinsey.com. Retrieved October 15, 2019, from https://www.mckinsey.com/business- functions/mckinsey-analytics/our-insights/what-ai-can-and-cant-do-yet-for-your-business 11. Cohen, T. (2019), "How to leverage artificial intelligence to meet your diversity goals", Strategic HR Review, Vol. 18 No. 2, pp. 62-65. https://doi-org.proxy.library.cornell.edu/10.1108/SHR-12-2018-0105 12. Eubanks, B. (2017).Using artificial intelligence for talent acquisition. Employee Benefit Adviser (Online), Retrieved from https://search.proquest.com/docview/1975083456?accountid=10267 13. Rodney, H., Valaskova, K., & Durana, P. (2019). The Artificial Intelligence Recruitment Process: How Technological Advancements Have Reshaped Job Application and Selection Practices. Psychosociological Issues in Human Resource Management,7(1), 42–47. https://doi- org.proxy.library.cornell.edu/10.22381/PIHRM7120194 14. Tambe, P., Cappelli, P., & Yakubovich, V. (2019). Artificial Intelligence in Human Resources Management: Challenges and a Path Forward. California Management Review,61(4), 15–42. https://doi- org.proxy.library.cornell.edu/10.1177/0008125619867910 15. Zielinski, Dave (2019). Separating Artificial Intelligence Myths from Reality. Society of Human Resource Management (SHRM). Retrieved October 2, 2019, from https://www.shrm.org/resourcesandtools/hr- topics/technology/pages/separating-artificial-intelligence-myths-reality.aspx APPENDICES Appendix A: Table I Areas AI tools can be employed to support R&S Ai tool Problem Solution Outcomes Adoption Vendors Vacancy Spontaneous Software identifies Improved talent attrition Large companies Workday talent prediction resignations increase employees’ behavioural Improved employer (e.g. IBM) insights software costs data and makes a brand Data-driven firms Bamboo HR prediction on likeliness Reduced time toh ire (e.g. Facebook) Job rate to leave High candidate Monster talent Prediction software volume (e.g. management gives a head start, Goldman Sachs) which reduces these High turnover (e.g. costs Call Centres) Job description Complex jargon, Software provides Improved diversity Cisco Textio optimisation boring, indirect recommendations to Reduces the risk of American Express Three sourcing Software discrimination can be optimise job indirect discrimination Johnson & Johnson 15Five off-putting descriptions and tailor Higher candidate Nvidia Negatively affects the language to engagement Expedia diversity, applicant different types of Evernote volumes and employer candidates brand Targeted job Wrong message to the Using AI, ML and data Improves candidate Retail sector ClickIQ advertising wrong audience insights, firms can experience Newton PandoLogic optimisation through the wrong target accurate Maximises chances of Netflix Recruitz channels is a waste of recommendations to candidate engagement YouTube Appcast resources relevant candidates Minimises advertising spend Multi-database Untapped potential of AI-tool scans through Accelerates candidate Intel Hiretual Pro candidate suitable passive multiple databases sourcing rate eBay Ideal sourcing candidates and former (e.g. LinkedIn, Frees up recruiter’s time Hilton employees reduces Glassdoor, indeed, to focus on more Verizon talent pool quality social media profiles) essential tasks IBM much faster and more Improves quality and Accenture accurately than a quantity of talent pool Warner Bros human recruiter CV Screening Reviewing CVs is time- Software instantly Reduces bias and issues IBM IBM Kenexa Software consuming and costly reviews a large volume associated with human LinkedIn Ideal. Human error increases of CVs to filter out and fatigue Hilton CVViZ as the number of CVs rank the best ones Improves diversity Goldman Sachs Zoho Recruit increases Reduces costs Amazon Talent Recruit Allows recruiters to focus Talent Cube on more essential tasks AI-Powered Outdated, boring and Tests use AI to provide Allows recruiters to focus Unilever Arctic Shores psychometric unengaging tests leads engaging tests on more essential tasks PwC Pymetrics testing to negative candidate designed to improve Improves diversity in the Accenture Knack experience and candidate experience work places LinkedIn negatively affects while simultaneously Improves the candidate Tesla employer brand assessing candidates to hire (C2H) ratio Video screening Pre-screening Software analyses Reduces bias and Vodafone HireVue software interviews are costly, video interviews to discrimination Intel Montage biased and time- assess person- Allows recruiters to focus Urban Outfitters Wepow consuming organisation and on other essential tasks IBM Interview St rea m person-job fit Improves candidate Hilton experience Unilever AI-Powered Background checking AI software scans Allows recruiters to focus Fortune 500 firms Check’s background is time-consuming and through multiple on more essential tasks Financial Firms Intelligo checking ripe with human error databases to verify Reduces costs Uber GoodHire Leads to problematic candidate details such associated with human Axa Insurance HireRight employee termination as criminal record, errors BT Sterling Talent downstream credit rating and McAfee Onfido references (continued) Table I Ai tool Problem Solution Outcomes Adoption Vendors Employer Reputation affects the Software scans through Stronger employer brand McKinsey & Co Lexalytics branding way candidates public data to assess improves talent pool Oracle Semantria monitoring perceive a potential overall sentiment and quality HP Microsoft employer identify weak points in Positive image for clients Dominos Thematic Bad reputation leads to the hiring process Reduces T2H, staff DiscoverText lower talent pool quality turnover and overall costs Candidate Direct recruiting and Chatbots are tool that Reduces T2H Sephora IBM engagement relationship leverages Natural Allows recruiters to focus eBay Nuance chatbot/CRM management are costly Language Processing on more essential tasks H&M NextIT and time-consuming to mimic human Improves candidate Pizza Hut Kore Unpredictable or high conversational abilities experience and Burberry Inbenta volume can lead to and can be used to employer brand Personetics longer responses, engage candidates, Aivi dissatisfied candidates, provide quick Mya which negatively responses to questions Beamery impacts employer anytime brand Automated Scheduling calls, tests, AI system that picks up Allows recruiters to focus AT&T X.ai scheduling interviews or meetings on scheduling on more essential tasks Disney Troops is time-consuming and expressions to Coca-Cola Tact non-essential automatically execute Walmart Olono these admin tasks General Electric Survey Monkey Appendix B. Use of AI Applications for Recruitment and Selection Appendix C. presentation from Gartner, 2018 Appendix D. presentation from Deloitte, 2017 Appendix E. Oracle survey results on The Utility of Specific AI-Related Tools for Recruitment Appendix F. Table 2 Where artificial intelligence can be most useful (%) Sourcing candidates 68 Screening candidates 57 Nurturing candidates 43 Scheduling interviews 39 Engaging with candidates 24 Interviewing candidates 17 Key benefits of artificial intelligence (%) 73 Saves time 49 Removes human bias 24 Delivers best candidate matches 34 Saves money Sources: LinkedIn; Statista; my survey among 2,400 individuals conducted November 2018. Appendix G. Table 3 Adoption of specific artificial intelligence use cases, by category (%) Current artificial All intelligence respondents adopters Sales and marketing lead scoring 64 87 Sales opportunity scoring 62 76 Sales forecasting 57 83 Customer service case 60 82 classification/routing Chatbots for customer service or product 49 79 selection Cross-selling and upselling 44 71 Fraud detection 54 67 Credit risk scoring 56 66 Email marketing 77 89 Appendix H. Table 1 Use of artificial intelligence and automation in interviewing and assessment of candidates (%) Increase retention 64 Evaluate skill gaps 61 Build better offers 54 Understand candidate wants 52 Do workforce planning 47 Predict candidate success 44 Assess talent supply and demand 42 Compare talent metrics to competitors’ 39 Forecast hiring demands 37 Sources: LinkedIn; our survey among 3,700 individuals conducted November 2018. Appendix I. Table 2 Potential impact of artificial intelligence on real gross value added, by country (by 2035, %) Baseline Artificial intelligence steady state United States 2.8 4.9 Finland 2.4 4.6 United Kingdom 2.6 4.2 Sweden 1.9 3.9 Netherlands 1.7 3.4 Germany 1.6 3.3 Austria 1.5 3.2 France 1.9 3.1 Japan 1.2 2.9 Belgium 1.7 2.8 Spain 1.9 2.7 Italy 1.2 2.1 Sources: Statista; our 2018 data