3 things to know when evaluating AI recruiting software

January 24, 2022

 

iCIMS Staff

7 min read

Artificial intelligence has permeated our lives in so many ways that we probably haven’t even noticed. A great example is a song called Not Easy, written with the help of IBM’s Watson AI, which recommended musical elements and lyrics based on a variety of different data inputs. For some people, it showcased the power of artificial intelligence. For others, it begged the question, “Will AI replace musicians? What if all music starts to sound the same?”  

Talent acquisition professionals may share similar fears about AI recruiting software. “Will it take over my job?” “What risks does AI pose to initiatives like DEI?”  

According to iCIMS Senior Product Marketing Manager Gazmend Kalicovic, these fears are largely unfounded. AI recruiting software has come a long way in many good ways. When implemented correctly, it has the potential to make life as a recruiter or TA professional a whole lot easier.  

At INSPIRE 2021, Gazmend spoke with Andreea Wade, Senior Portfolio Director of AI/ML at iCIMS, and Job Simon, Technology Advisor, about how AI can be used to advance the best talent without removing that essential human element. To hear the full conversation, watch their session here.  

Takeaway 1: AI recruiting software can streamline tedious tasks  

Recruiters are not being replaced. Instead, the objective is to help professionals be more productive and make more informed decisions. Think of AI as “applied intelligence.” Its purpose is to enhance human capability, not replace it.  

To better understand how AI supports talent acquisition today, let’s look at a few examples of automation.  

1. Automatic review of resumes. Many companies sit on a gold mine of hundreds or thousands of resumes that represent top talent. Recruiters often don’t have the ability to review each one manually. AI can automatically review 100% of available resumes, allowing recruiters to focus on more critical tasks.  

1. Prioritization of best-match candidates. AI can help you find that needle in a haystack for specialized or hard to fill roles by finding profiles in your talent pool that are similar to other well-suited candidates or your high performing employees. You can feel confident that AI facilitates your discovery process, not dictates it, through clear descriptions of candidate recommendation logic.  

1. Engage talent 24/7. Even when your recruiters aren’t online, conversational AI allows candidates to engage with your brand 24/7. Candidates can self-serve with a career site recruiting chatbot that answers questions, recommends jobs, and even schedules interviews.    

These are just a few examples of how AI recruiting software can be used in talent acquisition. As you can see, these applications help streamline recruiter workflows, elevating the role of a recruiter to be less administrative and more proactive.  

The bottom line: AI doesn’t replace human experience or instinct; it simply enhances it.  

Takeaway 2: Data selection is a crucial step in bias mitigation  

Another big concern around AI, specifically in talent acquisition, is related to bias. Is there any way to ensure that bias is completely removed from the hiring process? 

The answer is not that simple. While we all want to avoid building products that create exclusionary experiences and discriminatory practices, it is possible to inadvertently introduce bias during the data preparation stage.  

Data preparation involves selecting the attributes you want the algorithm to consider or ignore, like gender, education, years of experience, etc. Data selection is human-led which means bias can creep in despite best intentions. 

To help remove bias from the hiring process, businesses that use AI should be intentional about how they select the data used to make predictions. This AI playbook from The Center for Equity, Gender and Leadership at the Haas School of Business (University of California, Berkeley) recommends asking questions like:  

• Who benefits from the data collected?  
• Are different populations sufficiently and accurately represented?  
• Are datasets being appropriated for uses they may not be built or suited for?  

Additionally, it’s important to know that the model is an actively learning system. You can monitor the outcomes of your model to better understand how it is using your data to make predictions. Then, feed it back with the right set of information to improve its accuracy.  

As Wade says in iCIMS whitepaper, 4 steps to strengthen responsible hiring with AI, “The whole world of fairness and bias starts with data.”  

Takeaway 3: AI recruiting software can be trusted, but not blindly 

Many TA professionals wonder, “Can AI really be trusted?” The answer is yes, although you shouldn’t trust it without understanding the ethos behind the technology. Businesses can do several things to improve the trustworthiness of their AI recruiting software.  

Here’s how iCIMS approaches this matter:  

First, we have what we call our Ensemble AI. It’s our patent-pending approach that leverages three different engines, a multitude of algorithms, and a built-in polling system to make recommendations that reduce the likelihood of bias or error. As Wade says, “Three heads are always better than one.”  

Second, iCIMS is constantly training and retraining our algorithms. We add new data points several times a year from multiple resources, geographic areas, and so on.  

Third, we have a responsible AI program and a code of ethics that informs our bias detection and mitigation work. An international organization called AI Global, working together with the World Economic Forum, continually tracks and assesses major responsible AI initiatives. Through this work, they’ve created a unified framework called the responsible AI Trust Index. iCIMS’ code of ethics is aligned to this and its pillars.   

The role of AI recruiting software 

AI recruiting software isn’t a replacement for human intelligence. TA decisions begin and end with humans, recommendations from artificial intelligence are meant to better inform those decisions.  

Go to our on-demand library to watch the full INSPIRE session, from Artificial to Applied Intelligence: The Next Step for AI in TA.