The idea to create a more diverse and inclusive (D&I) workforce is not new for organizations. A series of high-profile and costly lawsuits in the 1990s and 2000s, primarily around sex scandals and gender discrimination, forced them to take the diversity issue more seriously.

Organizations had rolled out several initiatives to tackle the problem, seeking to create awareness among their workforce with various initiatives, but, most of these did not deliver the desired results, for different reasons.

However, a lot has happened on the social front in the last few years. From the Black Lives Matter movement to the mass adoption of remote work, leaders are rediscovering the importance of diversity and inclusion more than ever before. Organizations have begun to reflect on issues pertaining to social justice, equity, flexible working, privilege, and the importance of D&I.

The modern-day D&I initiatives are taking a holistic and long-term approach, and some organizations are making D&I their primary HR objective. Today's HR leaders also have access to a superpower called Artificial Intelligence (AI), which thrives on big data and can generate insights into latent trends in employee behavior.

Organizations are using machine learning and advanced analytics on unstructured data such as resumes, job descriptions, survey responses, etc., to inform HR leaders about the mood and perception of their employees. Can AI help HR leaders and business partners with their D&I initiatives?

People Matters interviewed Aswini Thota, Data Scientist at Bose Corporation, America's leading audio manufacturing company, to discuss how companies can incorporate AI into their D&I efforts.

Here are some excerpts.

Why is it important for HR to embed AI into their analytics strategy?

Until a decade ago, AI was seen as some shiny and futuristic technology that’s only used in research labs. In a very short time, AI matured so much that companies have begun to embed it in their mainstream applications.  Accessible and economical cloud computing coupled with the availability of big data has provided organizations with unprecedented access to powerful AI algorithms.

We have seen a meteoric rise in AI adaptation across the core enterprise functions such as sales, marketing, supply chain, and finance. HR is no different. AI turned out to be a good match for use cases such as attrition modeling, quality of hires, candidate matching, etc.

Predictive and prescriptive analytics will help HR develop proactive approaches when handling challenges in their essential functions. Advanced analytics-based decision-making will also enable HR to minimize biases and think strategically.

What makes D&I initiative so significantly different? How can organizations effectively leverage data analytics?

Many industry leaders strive for diversity and inclusion because it’s the right thing to do. A renewed understanding of an organization’s social responsibility is not only the reason organizations are focusing on D&I,  there are tangible economic reasons as well.

Studies have shown that companies that run more diverse and inclusive workplaces outperform those that don’t. A survey by Diversity Inc. found that employees who feel genuinely welcome in their workplace take 75% fewer days and exhibit a lower turnover risk of 50%, saving recruitment costs and increasing productivity.

However, implementing successful D&I programs is easier said than done. It takes meticulous planning and careful tracking. The D&I needle moves very slowly; it will take months or even years to see some tangible results.

This is precisely when data analytics can help you. It can help you identify the direction you want to take, to stay the course, and over time, analytics can even help you predict the upcoming potholes in your journey. In other words, it can help you with defining the diversity, creating metrics, tracking the KPIs, predicting and prescribing solutions.

Defining diversity: Defining diversity is not at all trivial. In fact, diversity could mean different things to different companies, and it all depends on the organization’s long-term objectives. The most common way to represent diversity is by gender, age, race, ethnicity, and language. This phase involves creating the logic to identify your current diversity levels anonymously. It’s essential to follow all possible ethical, legal, and compliance guidelines here. If taken loosely, it might backfire and create unwanted mistrust among the workforce.

Creating KPIs: Once we define what diversity means to us, it’s time to move on to creating key performance indicators (KPIs). External labor market sources such as Bureau of Labor Statistics, Ministry of Statistics and Programme Implementation etc., can provide reliable information about the external labor market by age, sex, race, ethnicity, etc, and these should serve as a benchmark for D&I initiatives, while KPIs provide a birds-eye view into D&I programs.

Advanced analytics and AI: Once we get the foundational understanding of our D&I programs, we can leverage advanced analytics, machine learning, and AI to be more predictive and prescriptive about the outcomes. The intention of leveraging AI is to take a more proactive approach towards D&I. AI will help you forecast the occurrence of events and provide you with valuable lead time to course correct.

What are some HR functional areas that can incorporate AI to improve D&I?

I want to highlight that AI for HR will yield powerful results when used to assist HR professionals. I firmly believe that completely autonomous AI solutions are not a good fit for the HR business unit; instead, AI should be used to discover hidden patterns and bring forward insights.

AI-driven solutions to improve D&I are primarily used by talent acquisition, compensation, and employee experience functions. Some use cases are strategic and will yield results in the long term, while others are more tactical.

Studies have shown time and again that women are underpaid for the same job as men. How can AI help with compensation bias?

Insufficient compensation has been identified as one of the top reasons for employees quitting their jobs - employees who left their jobs last year made overwhelmingly clear to their employers, and both US and India statistics bear this out.  According to United States Bureau of Labor Statistics data, in 2020, women earn 82 cents for every dollar a man earns. ADP's recent survey indicates that in India, only 65% of women received a pay increment or bonus for taking on a new role compared to 70% for men.

Compensation discussions in many companies are based on the manager's feedback coupled with the market survey data, organizations purchase from third-party vendors. The pay recommendations are often based on standardized dimensions such as job level, talent availability, cost of living, education, etc.

But, what about the skills you developed, customer satisfaction you provided, and institutional knowledge you possess? Traditional pay recommendations ignore or consider these attributes as intangibles. This vintage approach towards compensation severely hurts women more than men, as the external data you are using is only going to multiply the hidden biases.

AI-assisted homegrown pay recommendation systems will help companies generate personalized salary bands, the compensation that works for your employees. This means that you have the opportunity to change from a role-based to a skill-based approach, valuing the skills your organization needs. 

Can you give us a couple of practical AI use cases to source and hire diverse candidates?

Sourcing and Talent Acquisition teams are perfectly positioned to attract diverse candidates. I want to highlight a couple of projects I’ve implemented in the past.

Job descriptions

Social scientists have discovered that coded gendered language in job ads promotes and sustains gender inequality. For example, job postings in male-dominated fields like engineering and computer science use masculine-coded languages like code and conquer in job descriptions. These kinds of words discourage women from applying to these jobs because they feel like they don’t belong.

AI-powered tools can help recruiters and hiring managers write gender-less job postings by recommending synonyms for gendered languages. I’ve also found that biases go far beyond specific words in my research. The number of skills you use, the way you ask for years of experience, the presence of phrases that indicate flexibility and work-life balance, etc., also significantly affect your ability to attract diverse candidates. We implemented a Natural Language Processing-based solution to identify the drivers of diversity in the job descriptions and used those findings to educate our recruiters and managers. This can help companies attract more balanced candidate groups allowing diversity and inclusion objectives to be met.

Recommending places to post the job

A major paradigm shift precipitated by the COVID-19 pandemic is that many more employers are now willing to hire remote workers. This opportunity to hire talent that was previously beyond their reach in untapped job markets can not only increase a company’s diversity by hiring from a location with a higher percentage of diverse workers, but can also position remote employees to advocate for the company in their local area, improving brand awareness and sales.

The shift to remote or hybrid employees also gives organizations the option to hire less expensive talent, based on the local job market.

With this in mind, AI and machine learning can facilitate a company’s location and sourcing strategy. For instance, AI models can optimize and recommend the best locations to source talent based on key company metrics such as role, function, and seniority. This allows the sourcing team to avoid guesswork and reach out to qualified diverse candidates outside of the organization’s standard talent markets.

What is your advice to leaders who want to embed AI into their D&I strategy?

There is no doubt that advanced analytics and AI are here to stay... Though every HR process is an opportunity for AI, leaders should first prioritize use cases that can easily integrate with their current systems/processes. Doing so will help you to ease your way into taking up larger and more ambitious initiatives.

Additionally, when it comes to HR, business leaders and data scientists should spend extra time investigating the quality of the data and algorithms to ensure there aren’t any algorithmic biases.