There's no going back: how AI is transforming recruitment

The accelerated use of artificial intelligence and machine learning by recruitment specialists over the past year is creating jobs by the thousand; it's time for HR to fully embrace the new technology and work with it to avoid bias, argues AI expert Gez McGuire.

Long predicted to be a transformative aspect in all our lives, the adoption of AI and its applications expanded rapidly in 2020 in response to Covid-19. Advances that were expected to take years have been accomplished in months. With major firms already making use of automation in their recruitment processes, it's important to gain a picture of how the technology is already transforming who and how firms recruit, and the practical steps HR professionals can take to ensure they stay ahead of the curve.

In response to the pandemic, we have seen the adoption of new technological behaviours, from video-conferencing to remote working, reach levels that were not expected until 2025 or even 2030. As reported in The Economist, consultancy firm McKinsey acknowledges that recent data shows that we vaulted five years forward in digital adoption in both consumer and business behaviour in just eight weeks.

Improving recruitment

For some of the world's largest businesses, AI and digital adoption was becoming an essential consideration even before the pandemic. Vodafone is one such example. Each year the telecoms giant has more than 100,000 graduates applying for just 1,000 jobs. With such a high volume of candidates the company's HR department recently contracted with HireVue to test an AI application that removes human bias from the recruitment process.

The system works by extracting as many as 25,000 data points from video interviews. Examining visual and verbal cues while comparing word choice, facial movements, body language and tone to help identify the best candidates.

The programme also sorts the candidates into highly recommended, recommended and not recommended. Vodafone concluded that the AI system correlated well with its own internal assessments (around 70% for the "highly recommended" candidates).

While cautious because they are only at the early stages of this process, the company has stated that this is something that will be part of the future of recruitment. An extension of the HireVue project has seen Vodafone take their time-to-hire from 23 days to 11 days, reduce candidate dropout rates by 30% and triple cost savings.

What the Vodafone example shows is that when AI is trained properly with the right data sets, it can offer clear and tangible benefits, not just in terms of increased efficiency but also in candidate quality.

What roles will be needed?

AI isn't just transforming how businesses recruit, but also who they recruit, and how they support, develop and retrain existing staff. Global research and advisory firm Gartner predicts AI-related job creation will reach two million net-new jobs in the next few years. The World Economic Forum in its recent report went on to identify data analysts and scientists, AI and machine-learning specialists, big data specialists and digital marketing and strategy specialists as the top four roles seeing increased demand.

It may seem like it will only be those currently proficient in AI and data analysis that will be able to take advantage of the huge number of roles on offer. To reach the critical mass needed to meet with current and growing demand, a much larger recruitment base is needed. The good news is that the main requirement for those looking to retrain in AI is the same level of computer literacy found in most current roles. Furthermore, rather than taking three years to complete, many entry level AI courses take just 12 weeks.

Underpinning all of this is AI's need for data. AI is often seen through the lens of removing existing roles, a better interpretation would be a process of upskilling. Machine learning requires a huge amount of data, all of which needs labelling in order to be processed - something that by its nature requires human intervention.

Depending on the complexity of the AI model being built, the size of dataset required to train the AI through machine learning will vary from case to case. Other factors that influence this will be the performance that an AI model is expected to deliver. Often, machine learning practitioners will try to achieve the best results with the minimum amount of data or resources in order to build their predictive model. This will generally take the form of a simple model using few data points. Once a desired outcome has been achieved, they will then move onto building a more advanced model, and this is when the potential for a vast amount of data is required.

Avoiding bias

Time and again, one of the biggest flaws with AI applications, whether in recruitment or any other area, has been in-built bias. One of the latest examples was the 2020 exam fiasco in the UK when AI was used to predict student exam grades. The system got it wrong, but only because the data it was trained on was flawed and full of human bias. This is where teams of fully trained data analysts really demonstrate their worth, as not only can they work to prevent the programming of bias, but also allow organisations to modify AI programmes quickly and effectively when issues arise.

To stay ahead of the curve, HR managers need to approach AI from two sides. Firstly, to source the technologies and techniques that can be directly incorporated to improve elements such as the recruitment process itself. Secondly, to identify the skills each organisation is going to need in the next few years, in order to benefit from AI and machine learning, and what can be done now to ensure that existing talent is being developed.

For those who pro-actively train their existing staff, there is the added benefit of having individuals with an intricate understanding of the particular business sector they operate, enabling them to identify rapidly and adapt effectively to any in-built bias.

Change is now, and every business and organisation must adapt.