The Potential for Machine Learning in HR: 3 Business Application Examples
How machine learning can unlock the potential of your workforce
Business applications for machine learning are becoming increasingly integrated in organizations, and not just in the commonly known giants such as Google, Amazon, and Tesla. AI and machine learning are increasingly reaching midsize and even small companies, especially in operations/logistics, business development and finance/accounting domains. The speed at which companies are onboarding data science solutions is only accelerating, except for one field commonly left behind: HR.
HR is commonly perceived as a less “quantitative” business function, yet it’s probably the one holding the greatest amount of data (both structured and unstructured). Why?
There are definitely multiple reasons; the top three that come to mind are:
1. Data sensitivity and confidentiality concerns: HR data ranks as one of the most sensitive forms of information held by organizations. Concerns on who can access, view and edit are thankfully not taken lightly by most. Individuals with access rights are cautioned to use it only under necessary circumstances. This taints the image of data with a sense of distance and restriction.
2. Consolidation and reconciliation challenges: Data is often incomplete and the reconciliation across multiple sources can be complex, or simply time consuming. Due to security/confidentiality requirements, many companies are often found operating through multiple systems. Recruitment, health & safety, total rewards and employee record management are activities frequently led by four different teams operating in four different systems. Important consolidation and reconciliation efforts are required to obtain a comprehensive holistic view on the workforce.
3. Lack of data savviness amongst HR roles: HR does not sufficiently prioritize data analytics in its function. The extent is usually found amongst the HRIS (Human Resource Information Systems) sub-function of HR, but the role played by this team is often limited to user support for data extraction requests. In some companies, data accessibility is even limited to this sub function, widening the gap between HR professionals and the data.
Despite these three challenges, there are countless opportunities to leverage machine learning to manage and solve complex HR issues. Here are three examples for illustration:
· Diversity, equity and inclusion: A theme that is becoming even more critical over the last few years as employers are looking for ways to de-bias their companies and benefit from the strategic advantages of a diverse workforce. Unsupervised learning techniques can be leveraged to identify systemic biases within an organization. By using clustering techniques, the multiple data points captured on employee entry and movement within an organization can be used to support the identification of process or tools failures.
· Strategic workforce planning: The labor market is becoming more and more competitive due to the low unemployment rates, the increased demand for specialized talents, the gradual retirement of baby-boomers and the entry of a new generation of ambitious professionals with high expectations. Supervised techniques can be used to project short and long-term talent needs. Regression methods based on historical data (absences, turnover), internal workforce attribute (median ages, seniority), business needs (and their associated talent needs), competition, and macro economic factors (interest rates, employment volatility) can be assessed to quantify resourcing needs in the short and long term and implement the appropriate strategy with regards to internal mobility, talent management and recruitment.
· Employee experience: A company’s employer brand and value proposition plays an important role in candidate attraction and employee retention. Multiple machine learning techniques can be leveraged to better understand pain points in employee journeys and implement corrective action to elevate employee experience, based on existing captured data. For example, natural language processing algorithms can be used to assess qualitative data captured during performance evaluations or internal surveys/pulse checks to identify areas where the organization is vulnerable.
The best part of all these use cases? They do not require companies to change their existing tools or systems. These solutions can be coded based on organization’s systems and existing data format. The step to get there is smaller than perceived by most, but the gap with competitors becomes increasingly larger the longer companies wait to onboard these solutions, exposing themselves to losing their no.1 asset: their people.