In his recent article Decoding AI for HR, Bill Kutik explained that AI is an often wrongly-used, misunderstood term driven by hype rather than value. That is okay, it is part of the normal development curve for new technologies – for example: ‘big data’ or ‘gamification’. When something is new, we want it! We want to feel like we are at the bleeding edge of what is possible, even if what we are doing is building the same database as everyone else but just a little bigger and calling it ‘big data’. Or putting the same questions we have always asked in a speech bubble of a cute avatar and calling it a game.
The story of AI is no different. That isn’t to say we should dismiss anything that uses the term, nor should we jump on it thinking this artificial intelligence will now solve all our problems.
In hiring, AI is being used to describe different technologies. Let’s look at a few of them:
What it is: An artificial personality that help candidates (or recruiters) go through the hiring process.
What it really is: This technology is usually a combination of verbal language recognition or typed content recognition and pre-programmed response that are given either through a voice (think Alexa or Siri) or typed response (think chatbot).
This bot/assistant/friend will listen to or read the candidate’s responses and convert what it ‘understands’ into an input. From there, it will either populate that input into a field (voice to text like we use on our phones) or it will look up a response from a library and either say/type that response back to the candidate.
What it is not: This is not Rosie, Data, Robot, C3PO, Vicky [insert favorite beloved robot personality here]. It will not think, problem solve, connect with the candidate, or answer a question without a pre-programed answer. It also has limitations, for example, it may not understand what your candidates are saying/typing, or it could transcribe content incorrectly. And, it will likely not fool your applicants, they know they are not interacting with a person.
The value: Instead of thinking of this technology as AI (as it really isn’t fair to think C3PO and get Siri), think of this as a combination dictation technology and self-help interface. It can be a huge time savor for applicants to get answers to basic questions and get help filling out information as they drive home from work.
Solving: Answering basic questions and freeing up time for the HR team.
Doesn’t solve: Personal candidate engagement – they know it is a bot and not a person.
What it is: An automatic way of finding the best candidates – or filtering out the bad ones.
What it really is: These are algorithms that look for specific content and pass or filter out candidates based on that content. The content can be key words or the presence (or absence) of data from a source – such as a resume, form data, or profile.
Sometimes it is “intelligent” which means the algorithm is not created from a person specifying what they want, it is driven by a data set.
For example, if an algorithm used a data set that showed more successful engineers graduated from USC than UCLA it would filter out UCLA Bruins in favor of USC Trojans for engineering positions (I mean no offence to all the great Bruins reading this article – this example may have been influenced by the fact that I am a USC engineering graduate).
Sometimes the AI is “very intelligent” and includes ‘machine learning’ – which means that data set it is using changes over time as it collects more data.
What it is not: Intelligent. Yup, these intelligent Artificial Intelligent solutions are not actually intelligent. The AI cannot tell if the data set is good or bad – the data could be great… or it could be garbage. We all know what happens with garbage; what goes in reflects what comes out.
When looking at any AI selection solution – look very closely at where this “intelligence” is coming from. What data is the technology using to determine which candidates will pass and which will not? Just as important; look at how it determines how candidates meet that criteria. If it is exact key word match, for example, that is a lot of pressure for applicants to know the exact key word that the AI needs to see to pass.
Another area of caution is bias. These algorithms are made by humans, and we are biased.
For example, an algorithm that filters out anyone with gaps in their work history will unfairly filter out people who took time off to care for family, do mission-based work they did not put on their resume, lost their job due to an economic downturn, or any number of other perfectly valid reasons for gaps.
The value: Instead of thinking of this technology as AI (so we don’t have the unfair expectation of intelligence), think of this as a filter tool. It can be a huge time saver for cutting down the noise of candidates who will not fit your criteria. If it is drawing on a fantastic data set that shows strong, relevant trends on traits/attributes that are correlated to success, it could really help increase your chances of finding great candidates. However, be cautious: understand exactly what it is doing and make sure it is filtering on the right criteria and identifying that criteria in an effective way.
Solving: Filtering out the noise from the true applicants.
Not solving: Finding THE ONE. Filters will only get you so far. There is a very important human element to evaluating if talent will be a good fit for your organization. No algorithm is going to be as good as a human for that. It could also be introducing bias into your process and costing you great candidates if it is not filtering correctly. It will also not solve any issues with candidate engagement – candidates do not like being filtered out without feedback.
AI candidate sourcing
What it is: A candidate search tool that finds the perfect candidates rather than waiting for the candidates to come to you.
What it really is: Similar to AI selection tools, these tools use algorithms and candidate data sources (such as a job board and social media sites) to find matching candidates. Also, just like AI selection tools, AI sourcing tools are only as good as the algorithms and the candidate data sources. Bad algorithms or flawed sources will result in bad matches. Good algorithms and good data sources will result in good matches.
What it is not: A recruiter. These tools are great at seeing patterns, but they are not good at seeing the potential of individuals. They will not be able to tell if that gap in history is because of a family need, or that the pattern of switching jobs every three years is because the individual is married to a military officer, or that those perfect skills listed are actually a list of skills the person has heard of, not skills they have.
The value: This technology is great for creating a list for recruiters; identifying potential candidates who may not have found you or tapping into the passive candidate pool.
Solving: Increasing the candidate pool beyond active candidates who have already found your job.
Doesn’t solve: Replacing recruiters who can evaluate the potential of the individuals.
What to do about AI
Nothing. For now, consider AI just another fancy label trying to get your attention. Rather than selecting a tool because of its fancy label, choose what will help you get the job done.
Identify your challenges and needs: That way you are focusing on finding solutions that meet your needs.
Define success: To protect against finding a solution that fix one problem, just to create another.
Create a prioritized list of must-have and nice-to-have needs: Just as with candidates, if you define what you need for success up front, you can compare options and objectively choose the best one for you.
As you look for tools, consider this: No one buys a hammer expecting it to build a house. Don’t expect your talent acquisition tools to hand you your hires – tools just don’t work that way. At least, they wont until we have C3PO (or, if we are not so lucky, the Terminator).