Artificial Intelligence (AI) and robotics are pervasive in the hiring process. Built in to many Applicant Tracking Systems (ATS) that companies use, these technologies match open roles to candidates and weed out “mis-matches”. This all happens before candidates reach the “human” phone screen or interview stage and saves the company time.
As Dawn Graham notes in her Forbes article candidates can now be “disqualified in seconds” through AI and chat bots. At the other end of the spectrum a person in Sweden is believed to be the first hired by a robot with no human intervention.
Companies say bias is reduced or eliminated using AI in hiring because a human being is not involved in the up-front vetting process.
This is alarming.
Why? The way AI is used in hiring today delivers “bias at speed”. And contributes to a broken hiring process. In my research I spoke with recruiters and hiring managers alike who commonly said they could not find candidates with the right experience through these processes. Candidates told me they feel ATS is an impenetrable “HR wall” they can’t get through.
But AI itself is not the problem. There are more fundamental issues with the hiring process and its use of AI.
What are they?
The basic problem with the use of AI, chatbots and ATS in the hiring process is they are configured to work in the same way humans used to operate this process for eons. Just much faster.
What does an ATS do? Many ATS automatically identify candidates who should progress to phone screening or an interview by using AI to match the posted job description with the resumes of candidates who have applied. ATS are commonly set to present to the recruiter those resumes that have an “80% match” to the job description.
Why is this a problem?
- The terms and job titles used in job descriptions and resumes are not standardized and often not comparable. “Head of sales” or “engagement manager” can mean quite different roles at one company versus another. But AI systems see them as standardized and comparable and use them for matching — giving rise to bad matches and worse, incorrect rejection of resumes.
- Job descriptions do a bad job of communicating the underlying business problem the company is trying to solve by the hire. They mostly document the objectives & responsibilities of the role, and often are written by someone who is not close to the underlying business problem and/or is using a standard job description template. Which contributes to chosen candidates not being successful in the actual role as they may not be the best person to solve the underlying business problem.
- Resumes are a history of past roles but do a bad job of communicating a candidate’s unique value and the difference they could make at a company. This is particularly an issue when a candidate is pivoting from one type of role to another to leverage their unique underlying strengths in a new way — because most ATS look for similar past roles as an indicator of a candidate’s suitability. This problem will get worse as the nature of roles at organizations changes more rapidly as new technologies are introduced.
- Knowing how ATS work, candidates now tailor (“massage”) their resume to include key words and job title names mentioned in the job description to increase the chances their resume will be matched by the ATS. Apps are now available that simulate the matching done by an ATS and suggest what keywords should be added/changed in your resume to increase its match chances. This often results in a heap of resumes presented to recruiters by the ATS but which are hard to tell apart.
- Matching a poor definition of the need (job description/job posting) with a poor description of the value proposition of a candidate (resume) results in poor matches presented to the recruiting manager.
- Recruiters devise ways to whittle down the list of resumes presented by the ATS to a manageable number to be phone screened and/or interviewed. Still having not met the person, the methods used to do this culling are at best arbitrary and at worst discriminatory, eg resumes with spelling or punctuation mistakes, without cover letters nor bios are put to one side. I have spoken to many job applicants who say they are frustrated that they have great experience to perform a role but can’t get to a human at the organization to communicate this.
- And perhaps even more fundamentally, ATS only presents candidates who have applied for the role. What happens when the best candidates out there may not have applied?
Do you get where I am going here? Yes, this process sounds broken because it is! But it does not stop there…
- Companies often filter out candidates that do not have a college degree or have not filled in their year of graduation in the ATS. Many companies feel a college degree is a necessary qualification to be valuable to the company in any role. This is regardless of the value that a person might actually represent to the company.
- There are a significant number of companies which feel that older people or those at “retirement age” are not a good “cultural fit” for the company or will be too expensive to hire. Many older candidates try to avoid specifying their year of graduation in the ATS, or they exclude roles they had beyond 15 or 20 years back, worrying that this will indicate their age and feed the significant ageism that exists in today’s hiring. Ask anyone who has applied for a job online at age 45 or 50 or more how hard it is to break through the “HR wall”. Companies miss out on the value a person can offer who has the emotional intelligence and wisdom that often comes from deep experience.
[The requirement to specify year of graduation should be removed from all ATS immediately to stop feeding age discrimination!!].
What is the solution? What is a better way?
Let’s do away with job descriptions, job postings, and job applications with resumes, all part of the current hiring process which recruiters, hiring managers and candidates alike see as broken.
Instead let’s approach it this way…
Networking is seen as a much more successful way for a candidate to find a role compared to applying to a job board. [I discuss the value of networking in an earlier article.]
So why not use the process of networking to underpin our hiring systems?
What would that look like?
A matchmaking process. Where the operating model for hiring looks more like a “business” version of Match.com or eHarmony than a traditional job board.
The way this comes to life would look like this:
- Talent (candidates) maintain a “profile” of their skills, abilities, value proposition, and strengths on a platform. Part of their posted value proposition includes a description of the problems they solve at organizations. They maintain this profile even if not currently seeking a new role.
- Recruiters & hiring managers search the platform by entering some key parameters including a description of the business problem they are trying to solve.
- AI is used employing natural language processing to match the problem being solved to the people who solve that problem. So instead of matching “I need a CMO” to “I am a CMO” (old process) the AI would match “my sales are not growing” or “I can’t get lead generation working reliably” to a person who solves those problems.
- Other data points will of course be relevant to the matching like whether the role is full time or part time, methods of compensation sought versus offered, location of the role, whether remote working is OK, etc.
- Anonymity will be preserved in that candidates’ identities are not included in the match results. The focus is all about the skills, experience and value the candidate can offer. Candidates are alerted when they have been matched but the candidate must agree before their contact information is conveyed to the hiring manager/recruiter.
- This approach also reduces the risk of inherent and unintended bias in the hiring process. The hiring manager/recruiter is focussed at first on the value proposition of the matched candidate because that is all they have in front of them. No photo, no demographic information, no name — just the value proposition the person offers.
[Our brains automatically and unconsciously make inferences about a person before we have met them or gotten to know them, based on what we initially “see”. This inbuilt mechanism developed in humans over thousands of years to protect us during encounters with others we don’t yet know. But today it can be a driver of inherent and unintended bias. If we gain our initial impression of a person based on the value they could offer to the organization, before knowing what they look like, what their name is, their apparent age, ethnicity, gender, etc. then that should increase the chance that the best candidate is considered for the role, before we learn about the person’s other attributes. Especially because those other attributes, except in rare cases, are not relevant at all to the value they offer.]
I have tested this approach manually by matching “the problem the company has” to the “problems I solve” with 15+ companies I have worked with to find candidates for roles. In a number of cases I turned around the original ask of the company, for example “find me commission based sales people”, to instead understanding what the underlying problem was such as lead generation, and then finding people who solve lead generation problems. A common response was that I found the person who the company really needed, not who they asked me for.
So I know I am onto something here!
When executed on a platform leveraging AI, natural language processing and machine learning this process will drastically reduce the time it takes to identify great candidates for a role.
So let’s bring back democracy and a level playing field to hiring. And in the process have our organization’s problems solved more quickly…