Thank You Amazon for Messing up AI in Recruiting for Everyone.
AI isn’t perfect. It is still in its early stages and learning. It is also already showing great value to organizations that are using it to reduce bias.
Unfortunately, since Amazon’s use of AI went down in a flaming ball of fire talking about how biased it was you have a bunch of people not seeing past some of the sensationalized headlines “AI IS BIASED” to understand some of what actually happened here. As a result you have a bunch of articles, tweets and linkedin comments from practitioners and public that seem to agree on one thing “STOP USING AI FOR HIRING IMMEDIATELY”
As a result, we may see backlash against AI in recruiting and other areas of technology – especially among those companies that are highly risk averse or have low technical competency in hr – resulting in a slowing of sales of these products. This is an issue for a rapidly growing segment of the market that is trying to improve the experience for both the candidates and the companies and in part are doing things to help reduce bias. It wouldn’t be the first time that bad PR or an acquisition slowed adoption of technology down in this space, but it would be one of the most widespread given AI is now becoming a key factor in nearly every HR Technology.
An unfortunate saving grace for AI may just be an economic downturn that could occur late 2019 or early 2020. It has been pretty clear as an analyst we are due for some major consolidation in the hr tech market and other indicators and making this very reminiscent to the the other cycles we’ve gone through. If this downturn happens and we see a reduction in workforce, it may be enough to give AI the boost as it did with ATS in early 2000’s and video interviewing and sourcing tools in early 2010’s The challenge will be on the vendors using AI to either focus on the business outcomes and process it supports or educate hr teams on AI and show validation and ongoing monitoring – not just have splashy marketing featuring the word AI – something they haven’t had to do to date.
Some other things we need to keep in mind about AI and Bias in General:
First, People are afraid of change in technology. Especially that which they don’t understand.
AI is scary, I get it. I have been working in HR Technology for nearly 20 years. When we started introducing online job boards, people were afraid of their privacy. When we rolled out “online” ATS’s recruiters and human resources teams were hesitant and didn’t like it. Don’t even get me started on the drama around video interviewing and “seeing” a candidate. Social…same thing – resistance is ALWAYS part of the evolution of technology. Especially in HR. AI and Machine Learning is no different, it is just the latest in a long line of tech fear that we forget we had. What make AI in recruiting really scary for some is that it takes the “human” aspect out of it (you know, the one that has had proven unconscious bias and finite time to review resumes so many are just ignored) because…ROBOTS WILL TAKE MY JOB! At Accelir, my research has showed VP of HR and CHRO’s were often the most resistant to new technology and least willing to try to learn if they didn’t fully understand it or it seemed complicated. AI is complicated because it seems futuristic.
Second, It is important to note that Amazon Built This In House.
There is a saying “pride comes before the fall” – I always think of that as I hear from companies that attempt to build recruiting and talent management products in house. Why? Because I also get the calls on the other side when the project is a nightmare and they don’t know what to do or it has taken 3x as long as they were told. Just because you *can* build something doesn’t mean you should. Believe it or not, not even Amazon knows everything better than everyone else. There are HR Technology products on the market that have spent MILLIONS in validation and research around their AI to ensure they are not biased and then are able to continually monitor to ensure it doesn’t learn bias or can be caught if it does. The hard part isn’t building it – the hard part is building it with a full understanding of the nuances of hiring and where bias can sneak into the process AND continuing to validate and monitor for bias. I would guess most companies (including Amazon) didn’t have a full development team focused just on this, watching code and monitoring for bias consistently for the last few years.
Third, if your goal with AI is to replicate what you have – don’t.
If you have a goal when you buy technology to have it look for past trends to replicate vs using it to identify top candidates based on competencies and skill sets – you may have a serious issue. Odds are, your past actions are not reflective of what you (should) want your future outcomes to be. If what is “good” is the past 10 years of mostly white, male applicants from the right schools with the right degree then what is “bad” are women’s colleges or those those with non-traditional paths. The great AI based hiring technology I am seeing doesn’t look at things like a college name and looks at more competencies, skills or abilities.
Finally, Look at your Analytics Regularly
In any type of process – manual, automated, ai – look at your metrics and analytics on a regular basis. Not every few years – quarterly ideally but monthly or weekly even if you are high volume. If you are finding that the diversity or underrepresented candidates are not where they should be at any point in the funnel from sourcing to on-boarding are not representative of what they should be, identify and try to handle immediately. Even if you aren’t using AI and you are simply using assessments or creating candidate personas or profiles that look like your current teams vs your current teams skill sets and competencies – you are creating your own biased hiring and promotion practices within your company. No AI needed.