Using AI in Recruitment

We’ve decided that we’re going to use an AI to automate our tasks. This isn’t anything ground breaking, companies like Arya, Wade and Wendy have been developing AI for recruiters for years. We’re just using it to automate our workflow and cut costs.

So why AI? 52% of hiring managers say that the most difficult part of recruitment is sorting through a large applicant pool to find the right candidates. We feel the same way, we get hundreds of referrals and we only deliver the top 5 to clients. Some jobs get over 600 referrals and we’re stuck sorting them manually. We’re hoping to use AI to do smart resume analysis, technical interviews, sentiment analysis and other high volume tasks. The clear majority of candidates entered into the system are un qualified and thus the AI should be able to weed them out to leave us with only qualified candidates. The AI should then take these candidates through a screening process and in the end, select the best five and be able to spit out a report to the client to see. All of this could be done much faster as we’re no longer relying on human interviewers and relationship managers, the AI is awake 24 hours a day and is always available to work. The AI should also be able to standarise the process and remove human bias.

However, there are a lot of challenges in building a good AI for recruitment.

  1. Data. Training an AI involves a lot of data that isn’t always readily available. We have a data set that we can start the AI on but we’ll need more and more.
  2. Biases. Although I mentioned above that an AI should be able to remove biases from recruitment, it is still at risk of being biased based on the data set it’s trained on and the person doing the training. On the flip side, an AI might not be able to learn the tacit skills that we might need when talking/interview someone. It might not be possible to remove the human aspect entirely but at least we’ll lower the workload for the person.

Our AI will have the following skills:

  1. Intelligent Screening Software: the AI should be able to produce the best candidate given the information available. ML and AI will enable the system to learn more about candidates and job positions.
  2. Data Input: the AI will also be acting as a chat bot to input more information from both the candidates and the clients. If we require more information than what the candidate puts in, the AI will just ask questions as if it were a real person. This allows us to fill in any missing gaps in the data that we need quickly.
  3. Subject Matter Expert Interviews: we’ll train our AI to become expert interviewers, but more specifically to become experts in technical fields. We have a huge test bank of questions that candidates are using that we can feed into the system to allow it to learn right and wrong answers. We’ll also let the AI read transcripts from our current SME interviews to learn how to act the role of a technical recruiter. They can start by asking questions in during interviews they sit in on and eventually conduct interviews by themselves.

The use of AI and ML in recruiting is a great example of how computers and machines can help us with high volume tasks. What scares me a bit is how when I was writing the blog post above, I referred to our AI as if they were a real person without giving it a second thought.

We did come up with a name though.

World, get ready to meet AJ.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.