How AI can help you Recruit Faster - Better?
Course Lessons
S.No | Lesson Title |
---|---|
1 | Introduction |
2 | Let's look at a few key AI-driven features in Recruitment: |
3 | Conclusion |
Introduction
AI is doing some really cool stuff in the Recruitment / Talent Acquisition space helping automate some really tedious manual processes. In addition, there are some really good personalization features that it helps develop that phenomenally increase candidate and recruiter experience and enhance conversion rates and organisational branding towards the candidates and so on.
Let's look at a few key AI-driven features in Recruitment:
CV Parsing Automation: One of the first things that happen in the overall Recruitment process flow is an "Application" by the candidate against a job by submitting their "Resume or CV". Companies have two choices in which to structure the online Job application forms - 1) ask for all the details from the CV to be filled in the form, 2) just ask for the CV and parse information from the CV. Of course, for a better candidate experience, the second choice is better.
However, the challenge is in the way candidates write their CVs, their structure, format, level of detail, language, and so on are heavily inconsistent from candidate to candidate. Standard pattern-based document parsers are of little use here. Instead, AI-based Resume Parsers that leverage Natural Language Processing capabilities, especially through NER or Named Entity Recognition techniques can perform choice 2 with significant accuracy.
Smart AI-Based Search: Employers or companies providing hiring service end up having large databases of candidates or a large set of applications for their popular job posts. The challenge from that is how do you find the best candidate amongst say 500 applications? Manually, it's a humongous job. AI-based techniques can come into help. You already may have a large amount of data on previous positions and their screening and selection results. This data can then be used to train AI models to be able to come up with screening decisions through your system. There are pre-trained models also available that can be used.
Caveat on Ethical use of AI: If you are using your organizational data to train a Smart AI model for CV screening, then you need to ensure that the data (history of selections in your organization) are free from (human) bias. E.g. if managers in your company used to be more inclined to select from certain preferred demographics, then you are training your AI model with the same bias. A conscious debiasing effort must be taken up in this case.
Finding Similar CVs: It is possible to develop an AI model to semantically understand a document (Resume in this case) and find out semantically similar documents (other resumes). Texts (in Documents) can be converted into machine-understandable formats called Embedding Vectors. These Embedding Vectors combine semantic information that can be used to compare with other Vectors and scaler distances that reflect the similarity of documents. Using these techniques, you would be able to find Resumes that are similar to the one that you have selected.
Feed the JD and get matching CVs: This feature can also be developed using the Embedding techniques mentioned above. JD in this case is a document, CVs or Resumes are other documents that can be semantically matched using their respective vector representations to find similarity scores.
If you are into recruitment services, there are chances that you are already familiar with popular public portals and may have noticed some of these above features are already available. So, now you know how to develop them.
Send personalized Positions to candidates based on their search habits: Develop subscription features within your portal for candidates and capture their preferences either implicitly through their browsing data or explicitly by asking them. Once you have their Job preference information, you can develop AI-based features to find active positions that match. There are two choices - 1) you may use a standard text/word match-based matching system or 2) use a semantic search (AI-based) mechanism to find more reliable results for the candidates.
Share suggested articles about the company with the candidate: For this feature also, you may use Semantic search/comparison methods to find articles about your organization and send them to the candidates. This will help them remain engaged with your organization as a potential employer.
Share suggested profiles to Employers:? Once you post a job on your portal, you can develop a feature that immediately suggests potential candidates from your database. This can again be developed by semantically matching the JD and Candidate profiles.
The Semantic Search and Compare capabilities are standard NLP features today. There are multiple techniques available today, including BERT (Bidirectional Encoder Representations from Transformers), which is based on Transformers, a deep learning model. BERT-based APIs to convert your documents to Embedding Vectors and compare them resulting in scalers (numbers) can today be used very easily by AI developers.