top of page

Can Recruitment Technology fix hiring bias?

The working world is going through a rapid and extended period of transformation.

The number of companies snapping up talent and intensifying demand on the talent pool is increasing daily, proving that a big brand and paycheque are no longer enough to retain good people, especially those in high demand.

Recent figures from Gartner have shown that over 16M working-age people are currently excluded from the job market.

When layered against a backdrop of the great resignation, relentless rhetoric surrounding the war for talent and regular government updates of millions of open roles in the UK….

Something is not aligning.

We all know that having a diverse workforce has been proven time and time again to give an organisation the edge.

Winning clients, attracting talent, securing funding or even taking a company public can now depend on the diversity of your organisation. For example, Goldman Sachs recently announced it would no longer support initial public offerings of companies with all-male boards.

The returns on diversity are tangible, especially regarding innovation: a survey of 1,700 companies across eight countries found that organisations with above-average diversity had approximately 19% higher innovation revenues and 9% higher earnings before interest and tax margins.

(yes, there’s other more recent research than McKinsey available to support our business cases)

However, humans are inclined to bias, whether intentional or not. And as also found by Gartner, these biases almost always lead to hiring the wrong person, with the wrong skills, who is just as likely to leave within the first 12 months as they are to stay.

Heading into a recession, organisations are therefore looking to mitigate these significant risks and to make objective decisions regarding hiring, firing, promoting, and developing workers.

Whilst building the first generation of Clu over the past 24 months, we’ve analysed north of 500,000 job ads, spoken with almost 50,000 job seekers, and are running an ongoing qualitative study with hiring managers to understand where the breakpoints are in unlocking the entire talent pool and setting it up for success at work, equitably, in the process.

During this time, we have learnt so much about the future of recruitment, particularly around plugging the skills gap, sustaining talent pipelines and boosting retention. So, I thought I'd share some of our key findings with you.

Lesson 1: CV-sifting doesn't work. Period.

If you follow me on LinkedIn, you’d know I think the CV is one of the most dated and no longer fit-for-purpose tools used in hiring. Not to mention that the traditional process of scanning and tossing out CVs is riddled with unashamed bias and presents the biggest barrier to entry for diverse talent in the hiring process.

According to one intensive academic study, minority applicants who 'whitened' their resumes were more than twice as likely to receive calls for interviewers, and it did not matter whether the organisation claimed to value diversity or not.

Many vendors have now touted AI, well, more accurately, automation and digitisation, to overcome bias and boost representation via ‘smart’ CV sorting, but there are two problems with this.

Firstly, sorting CVs fundamentally misses the point of inclusive hiring. Currently, anyone can apply for any job. The impact on employer brands, psychological safety and talent attraction is palpable.

But ON TOP OF THIS, we spend millions each year telling people they should come and work in opportunities they don’t stand a chance of being hired for.

I always say you do you regarding employer brand budgets. But the most considerable burden on anyone is talent is all the irrelevant and unsuitable applications that land in their inboxes or ATS.

Every single other business service line has learnt that more is not more. And in recruitment’s case investing in broadening your funnel with any candidates, not the suitable candidates.

This is not only affecting the performance of your current machine learning hiring tools because they don’t learn correctly from the right data, but they also result in HUGE investments to filter the top of your funnel, taking vital funds away from people experience, L&D and retention programmes.

Secondly, automated CV sifting makes existing bias worse.

Consider Amazon's experimental recruiting engine. Having learned from the training data that Amazon’s male candidates had more experience in technical roles, the application started penalising resumes that included the word 'women'. (Amazon did wisely scrap that project, but the case remains.)

Nothing can save the CV in its current state, and algorithms are no panacea for D&I.

But when we apply technology within a thoughtful context, fully understanding how hiring inaccuracy occurs and can be reduced, we can start structuring enhanced data ontologies effectively, not efficiently and begin challenging legacy parameters with quantifiable analytics – not just “doing the right thing” business cases that may or may not be quoting McKinsey.

Lesson 2: Improving representation is more about Culture than Attraction.

Using bias-mitigating technology for candidate hiring and assessment is only part of the solution for attracting and maintaining a diverse workforce.

Ramping up efforts to increase the representation of certain groups will all be for nothing if these individuals are having a poor experience after they join the organisation.

Myth buster! This is why we also believe that the name-blind CV process will soon become a thing of the past, as spending millions of pounds telling the talent market how our employees can bring their whole selves to work and then removing people's identities in the hiring process is not just VERY TELLING, but it also shows a lack of intent to improve organisational accountability and transparency around bias.

Russel Reynolds found that only 35% of Chief Diversity Officers are tracking diversity data currently – and within that, this data is USUALLY limited to top-line race and gender demographics.

Suppose an Employer is not tracking this data or, worse, hiding results because of fear of shame (we see this more often than we'd like to admit). In that case, they will be missing essential markers that will help determine whether people are exposed to fair hiring practices.

By using better data and processes to help organisations build the kind of employment market where all talented people have an opportunity to thrive, regardless of their background, you inherently teach AI-led systems accountability for data irregularity, which is the most potent challenger to selection bias.

Lesson 3: AI can unearth a candidate's true potential.

AI might not be the best for increasing who can apply for jobs, but it can be used powerfully to support the closure of skills gaps in your organisation.

We spend more time than I'd care to account speaking about skills gaps but then completely erase skills-centricity from the earliest stage in the hiring process.

Many job seekers rich in transferable, behavioural and technical skills are merely exited from the process because an AI-led system doesn't recognise the correct keywords, years of experience or job titles in their CV.

By focusing on improved ontologies to identify skills and where and how they can manifest in and out of work, AI can elevate higher potential job seekers based on this and prioritise opportunities to those that can fill in the gaps in your organisations, thereby creating a more balanced and cognitively diverse workplace.

Being upfront and determining what skills are required to succeed in a specific role allows companies to massively evolve their recruitment strategy, enabling them to make far more informed decisions about a candidate, over and above gut feel of whether they're the right fit or not - again citing Gartner's shocking findings on why this is literally the worst thing you can do if you want to future proof your business.

This not only significantly scales the breadth of 'skills-aligned' talent that could be exposed to an opportunity but also significantly impacts removing barriers to entry.

This is why we've built Clu's dashboards around displaying what candidates can do, not what they've done to qualify them for processes.

This helps unearth a candidate's true potential and mitigates the risk of hiring based on what someone has done instead of focusing on what they're capable of.

Lesson 4: Pair AI with human expertise for transparency

There is a clear need to create a universally integrated job-matching tool powered by AI to prevent the billions of pounds wasted annually on unsuccessful hiring.

The most significant oversight in AI currently being used in recruitment technology is that it assumes candidates and Employers know what they're looking for. The focus on reach over accuracy deepens the trust deficit between talent and Employers, with larger organisations with the most expensive recruitment technology solutions stung the most.

We've found that the intentional application of machine learning at the point of role exposure for job seekers and posting for Employers can shift the quality of applications by up to 60%.

Helping Employers better understand what they're actually asking for and making success-based recommendations means that the jobs that end up in the marketplace are much more compatible with the most valuable skills a job seeker has to offer, and they are much more likely to find one another.

But none of this works, so long as we still cling rigidly to the old, reckless ways of working, which benefits neither job seekers nor Employers.

While other job families are rapidly adopting AI and machine learning, recruitment/HR relies on AI layered on top of processes that are proven not to work.

We have lived in the age of personalisation and micro-targeting for decades now. However, volume and reach are still the comfort mechanisms of how we deploy our most significant line item - ironically, also the one that underpins our competitive advantage.

Our vision for Clu is to build a hiring journey that sets every person who engages with it up for success by improving accuracy, experience and inclusion in the early stages of the hiring process.

In doing so, we shift the cost dependency of hiring from the top of the funnel to the bottom, meaning an Employer can save over 600 hours when compared to traditional methods of recruiting.

Transferable or human skills are going to remain the one area where humans will trump robots for many years to come.

Workplace culture, employee experience, diversity and inclusion are growing in importance, which is why it is more important than ever not to try and use technology to replace people but use the best parts of both to complement and hold each other more accountable.

To avoid biased output from AI tools and predictive analytics, when it comes to dealing with hiring, it is vital that the data collected is reviewed thoroughly to identify and remove any biases before using it. Also, samples used to train the machines should always be trialled and modelled regularly, and, if need be, algorithms should be adjusted accordingly by experts.

The future isn’t doom and gloom, though, because significant advancements are being made.

Despite hyper-critical press speaking about current solutions, we’ve found that when you remove CVs, improve data ontologies, focus on skills and increase accountability, not only does the performance of your hiring function thrive, but the demographics, engagement, culture and retention of your organisation do too.

Can we learn to create more ethical, inclusive hiring machines? If we get the balance right, absolutely.


For more information on how Clu uses machine learning to improve hiring channel performance and shatter barriers to the job market for systemically underutilised talent, get in touch with our team today.

Bình luận

bottom of page