The imminent ripples of influence from the AI revolution are being felt end to end across the entire spectrum of business management, and more so than ever within the Human Resource (HR) management sphere. McKinsey’s latest forecast predicts that AI’s impact on the world economy will generate up to $13 trillion in economic activity by 2030 on a global scale. Moreover, a 2017 survey by IBM of 6,000 executives titled, “Extending expertise: How cognitive computing is transforming HR and the employee experience”, found that 66% of CEOs believe that cognitive computing can drive significant value and transformations in HR management.
Business functions in which AI has been adopted, by industry, % of respondents1
AI seems to be gaining the most traction in the areas of the business that create the most value within a given industry.
1 This question was asked only of respondents who said their organisation have piloted or embedded at least 1 AI capability in 1 or more functions or business units. Respondents who answered “don’t know” or “none of the above” is not shown. For telecom n=77; for high tech n=215; for financial services, n=306; for professional services, n=221; for electric power and natural gas, n=54; for healthcare systems and services, n=67; for automative and assembly, n=120; for travel transport and logistics, n=55; for retail, n=46; and for pharma and medical products, n=65.
Value to date from AI adoption, by business function, % of respondents1
Across functions, respondants report that the most significant benefits come from adopting AI in manufacturing and in risk.
1 Respondents who answered “some value”, “no value”, or “don’t know” are not shown. This question was asked only about business functions where respondents say their organisations have deployed AI, and only includes responses from respondents who say their organisations have piloted or embedded AI in 1 or more functions or business units. For manufacturing, n=272; for risk, n=285; for supply-chain management, n=299; for product and/or service, n=536; for strategy and corporate finance, n=155; for service operations, n=669; for marketing and sales, n=482; and for human resources, n=198.
However, on the flip side, many organisations are still hesitant in allowing non-human entities to handle certain aspects of business processes, in part due to the pop-culture driven much-maligned concepts of AI.
The Human Resources Professional Association (HRPA) reported in a 2017 survey that 52% of respondents indicated that their respective businesses were unlikely to adopt AI in HR management in at least the next five years. About 36% believed that their organisation was too small to leverage its true power cost efficiently, while 28% said their senior management did not see the value in the need for such technology in the near future.
It isn’t as though AI will replace human employees altogether in some near utopian future, or dystopian if you see it that way, but it does have the potential to streamline the HR workload significantly. In the majority of use cases today, machine learning is used to lend efficiency to HR management as an enabler – meaning less time for HR practitioners being bogged down in administrative processes and more time for people management. In the case that this all sounds like science fiction, let’s explore a few examples of how AI is being used right now in HR.
AI Use Cases in HR
Personalised Employee Learning and Development
Given the changing nature of today’s workforce demographics driven by unparalleled levels of diversity, be cultural or educational, personalisation has become an integral part of attracting, onboarding and developing top talent. Companies like IBM are exploring how AI can effectively be woven into an employee’s onboarding process. They are currently developing a system that is designed to answer new employees’ most pressing questions in order to help get them up to speed fast.
Like IBM, businesses can now identify the unique needs of different individuals and recommend personalised training and development opportunities as well as rewards and recognition by utilising machine learning to develop recommender systems. Such systems are what you see on Netflix and Amazon and well, on any established retail or content/media sites these days. The more complex versions of the recommender systems implement unsupervised machine or deep learning clustering techniques to generate user personas based on the enormous data they have garnered over the years and assigning new and returning users to a given persona based on their browsing and purchasing behaviors.
Use cases that cannot leverage datasets of such massive scales, at least yet in most businesses, can implement similar recommender systems for new hires through either one of the two methods below:
• Collaborative Filtering:
Applying a basic machine learning classification algorithm such as K-Nearest Neighbors (k-NN), recommender systems can be implemented through collaborative filtering systems. A system such as this is solely based on the past interactions between users and the target items; for example, a history of an employee’s previous training courses. Hence, the input to a collaborative filtering system will be all historical data of user interactions with target items. The core idea behind such systems is that the historical data of a user should suffice to predict what their next needs are, i.e. no additional information is required other than historical data – no extra signals from the user and no presently trending information. However, one can quite easily imagine the realistic limitations of such a method, such as needing a past user history (effectively ruling this method out for new hires) as well as ignoring other critical parameters and signals in decision making.
• Content-based filtering:
Compared to collaborative filtering, the content-based filtering approach uses additional information about the user and the target items to make recommendations. For example, a content-based system might consider the educational background and age of an employee in addition to the history of past training courses taken, if available, to recommend other training courses or developmental opportunities. Content-based filtering systems are inherently more computationally expensive than their much simpler collaborative filtering system counterparts, yet significantly simpler than the unsupervised learning recommendation system algorithms used by Netflix, for example.
In today’s world of continuous learning, AI systems such as this can help by guide employees through learning and self-development opportunities beyond just the onboarding phase for new hires, accelerating workforce development for businesses. AI is now capable of building personalised learning paths through conversational analytics, leading learning and development to new horizons.
One of the more controversial use cases of AI in workforce analytics is that in employee retention – predicting when an employee could be heading for the exit door. As invasive as it rightly appears, AI bots can sift through employees’ logged information such as internet usage patterns, time spent in meetings, stock options withdrawn, etc. to identify anomalies from the established baseline that may suggest if and when they are planning to quit their roles. Additionally, with Natural Language Processing (NLP) making significant strides, AI can pick up on linguistic and semantic cues, and can alert employers when a change in tone is detected. There are a multitude of these “employee tracking” platforms, such as Veriato AI’s Vision, that are capable of this, but there are some obvious privacy concerns. It will be interesting to see how widespread the adoption of this particular use case becomes if the platforms’ capabilities are not bound by standardised privacy regulations.
That being said, companies spend a significant amount of resources in training and development of their employees. The capabilities of such platforms could alternatively be leveraged to detect underlying employee underperformance issues and help develop customised solutions for each employee, and find sources of intrinsic motivation.
Tied to the note on which we ended the potential use case of employee tracking platforms for retention detection, the same platforms can also allow HR departments to gauge employee engagement and job satisfaction more accurately – on a micro (individual employee) level as well as the macro (organisation or department wide) level. On the micro level, this is incredibly beneficial considering how important it is to understand the overall needs of employees; however, there are several key macro level benefits to having this information as well.
According to a recent report from the HRPA, these AI platforms can evaluate key indicators of employee engagement as well as success in order to identify those who should be promoted or could be engaged in a different vacancy, driving internal mobility. Doing so has the potential to significantly reduce talent acquisition costs, bolster employee retention rates and improve employee engagement.
Sentiment analysis using NLP allows HR teams to interpret potentially vast quantities of text data to uncover attitudes and potential areas of concerns from, say, staff surveys. What’s more is that the combination of AI-based NLP and machine learning allow the analysis of not just surveys but also of open-ended communications across an organization (e.g. across Slack and email). As NLP evolves, it is becoming increasingly effective at distinguishing between different types of emotion; even so far as being able to tell the difference between, say, expression of confusion and that of worry.
For your HR team, this type of AI-driven technology can provide a valuable early warning system. Being able to detecting rumblings of discontent on a particular issue as a macro level trend gives businesses the chance to respond proactively to issues before they becomes translate into larger problems. However, privacy concerns apply in this use case as well.
A simpler implementation of AI is that in the automation of a lot of the mundane, repetitive, and low value-added administrative HR tasks, allowing HR practitioners to focus on more strategic aspects of the work.
AI tools can automate common HR tasks like benefits management and handling common questions or requests by implementing well-established technologies like chatbots, which themselves are based on deep learning models involving neural networks. Additionally, it provides similar streamlined efficiencies during employee onboarding process such as preparing reporting reminders and scheduling interviews.
Recruiting: Talent Acquisition
A range of machine learning applications is already in the works in many top tech companies with the goal of improving their chances of attracting top talents in today’s competitive markets – be it competing with other tech giants or competing with the gig economy. Companies like Glassdoor have effectively used machine learning to narrow down searches to seek suitable candidates based on intelligent algorithms. These algorithms are another example of recommender systems – imagine LinkedIn’s job search feature recommending job-seekers positions that they are best suited for, but flipped on its head; this time the platform recommending candidates from its database to employers based on a vacancy they have posted. A similar machine learning application used to find and attract top talent is one developed by PhenomPeople, which seeks out prospective candidates for employers on a number of job platforms and social media and networking sites.
AI-driven recruitment tools such as Pomato provide a good example of the possibilities. It allows employers to upload a selection of CVs/resumes, define their desired employee skillsets and knowledge bases, and ask the tool to ‘read’ the documents and with some variation and combination of NLP algorithms they are able to recommend the most promising candidates to the employers. Besides saving time, it also prevents the involvements of some of the inevitable aspects of subjective human biases. By focusing only on the attributes that matter, AI-driven technology can help introduce the welcome element of objectivity.
Reduce Hiring Bias:
Humans are inherently biased. Even when striving for inclusiveness and promoting diversity, HR professionals may subconsciously lean toward a particular candidate; for instance, someone who is more like the recruiter or a person they know and admire and more likely to dismiss other candidates if they have unfamiliar attributes, including name. This is referred to as “unconscious bias”. A study conducted in the UK revealed that on average 20% of applicants of white British or European origin received a positive response from employers while only 12% of minority ethnic applicants, such as those of South American origin, who applied with identical resumes/CVs and cover letters, received positive responses.
Postive Responses, by applicants region of origin
AI algorithms can be designed to help employers identify and remove these, often unintended, biases. That potentially translates to better hiring communications, attracting a more diverse group of candidates as well as finding the right candidate for the role irrespective of their backgrounds. To put it in context, AI allows HR managers to go beyond gut feelings and rely on data-driven assessments. That being said, the dataset that the AI platforms are trained on, will also have to bias free. If the AI algorithm is being trained on a biased data, it too will deliver biased results, as was the case for Amazon, where its platform tended to favour male applicants only because of their gender.
The Way Forward
A recent Oracle study on advanced analytics in HR departments identifies the areas where AI is being used the most among respondents. Obviously, this study was only restricted to employers who already had a well-established AI framework in HR.
Current AI Use Cases
In fact, Deloitte’s 2019 Global Human Capital Trends survey found that only 6% of respondents believed that they had the best-in-class recruitment processes in terms of technology, while 81% believed that their organization’s processes were standard or sub-standard. The data shows that there are huge areas of opportunities for employers to adapt their processes and reap the benefits of using this AI to streamline HR management and workflow.
By no means is AI an all-or-nothing game. A more practical approach to its implementation will likely involve identifying specific HR workflow bottlenecks for your business and considering exactly what processes and tools will enable you to address them. For example, while you may be more than happy to rely on an analytical tool with NLP capabilities to help you narrow down a talent pool, you might still want to keep final stages of assessments firmly in the hands of certified and seasoned HR practitioners and recruiters.
At the end of the day, AI is an enabling tool. The automation of tasks through AI technology allows for the liberation of HR professionals from administrative tasks that do little justice to their skillsets, and instead empower them focus on uniquely human abilities such as critical thinking, creativity, and empathy towards HR strategy development and implementation. As much as the HR sphere continues to be disrupted by AI, HR professionals and AI solution providers must find ways to balance these advancements with transparency and protection of right to information privacy. Advances in AI technology will continue to change the way HR teams function, but it is already clear that with the proper implementation it can bring benefits in many folds any business.