I am not very fond of social engagements. I like to meet people but I don’t like the planning involved. However, I try really hard to go to parties, weddings or whatever social affairs I am invited to. But the primary reason is not partying, it is to find prospective employees. That’s how important recruitment is for us, constantly pushing us to the edge of our comfort zone.
Augustine Antony (Augie), is a senior leader at iarani Inc., who’s is primarily responsible for hiring. When I asked him to take it up, he was quite enthusiastic about it, like everything else. However, with time, I saw him sinking in the resume tsunami.
A few months back, the story changed. We started getting better candidates, ones that fit our culture. This is the story of change. How Machine Learning, particularly, Natural Language Processing transformed our hiring process and spared me awkwardness of parties.
******* Over to Augie *******
As a senior leader at a fast growing startup, I, Augie, wear many hats. Some days I focus on streamlining operations. On others, I am recruiting talent. I never liked my recruiter days because those days I could do nothing else. But then technology came to rescue.
Problems with Traditional Recruitment Process
Our conventional process began with posting a vacancy on the company website. Once posted, I would, then, visit various recruitment platforms such as Naukri.com, Monster Jobs, Indeed, LinkedIn etc and post the job there as well. Once the job goes live on these platforms, candidates would apply for that vacancy on these job platforms, like any standard recruitment process. A few days in, and my inbox would be flooded with links to resumes. The applications would go up to two to three thousand resumes per day, no joke.
Once I had enough applications to sift through (usually around the time my inbox resembled the spam folder), I would compile them in a spreadsheet to filter out the best candidates. Obviously, my work doesn’t stop there. Once compiled, I would then cross-reference their work samples online (if any) and finally narrow some ten to twenty thousand applicants into a concise list that would then be called for personal interviews, that meant emailing everyone individually.
It’s easy to infer that this process was far from being efficient. Moreover, it didn’t deliver the quality I had expected considering the number of hours I would spend in search of the ideal candidate.
In Search of Finding a Better Recruitment Process
It didn’t take long for me to conclude the inefficiency of this process. I found myself wondering what changes can be introduced to efficiently and effectively achieve a list of candidates that are fit for the job? How can we maintain the quality of potential applicants without putting in as much time?
We are a Digital Transformation Agency after all.
So one Beer Friday, I discussed the problem with one of our engineers; who after one too many drinks coded up with a centralised hiring platform.
He had created a simple form on our website through which applicants could directly apply on the website, after seeing our recruitment ads on job portals and clicking on the apply link.He had even made a provision of a personal dashboard, for me, where I could view all applicants on a single page without spending hours compiling them on an excel sheet.
The following sober Monday, more ideas flowed. We created filters and a resume parser (that completely eliminates the need to download resumes). While the technology simplified and sorted through the numerous submissions received, the machine learning feature ranked each candidate’s profile. This yielded better quality results. It also makes the resumes searchable.
It became our one stop shop for all my recruitment worries.
Extending the Recruitment Platform with Artificial Intelligence
Within a month over sifting through hundred thousand resumes, we realised we can extend this platform to process even more resumes and find the best candidate.
Here’s where we started using Natural Language Processing (NLP). I already told you that we had created a resume parser. Parsing is a technique where the system would read the contents of a document and organise the contents in a structured and an organised manner. With NLP, we could even eliminate the need to read structured data. NLP would categorise applicants based on relevance into respective buckets. This narrowed down our list of applicants who could then be pursued for personal interviews.
Automation to Ease My Pain
Pursuing candidates for personal interviews, via emails, was another pain point. It came with the hassle of keeping a track of the candidates that accepted, those who wanted to reschedule interview dates. On an average, we interview forty to fifty candidates per position. It takes a quarter to close a position!
I turned to my engineers, who automated this ‘hiring platform’. I was given a button to send automated interview emails. The platform automated and update schedules, reminders (to both me and the applicant) and even feedback, if asked. Candidates were also given a feature to reschedule their interview dates themselves.
Sadly, the machine doesn’t do it all. I would still have to interview the candidates, or so I thought.
Pre-Assessment Through One-Way Interviews
We introduced one-way interviews for pre-assessment of the applicants. We entered a set of questions, subject to the position they are applying for, into the platform. The applicant is sent a link where the applicant uses to answer the question the platform asks them. Since, this is in a video format, we were able to understand not only the candidate’s work merit but also their personality. We were successfully reaching decisions by being able to assess the candidate’s composure, speech, body language and his vision through these one-way interviews. This enabled us to really zero in on a select few and call them for personal interviews, saving us time and effort.
Personal interviews is probably the only manual effort I make when recruiting talent. After personal interviews (or even before), the aforementioned automation let’s me automate regret and confirmation emails to ensure due clarity is given to the applicant.
So far, this system has served us well and we are still evolving. If you can’t tell already inspiration strikes us best on Beer Fridays.
We often invite our clients to our Beer Fridays. One such Friday, our engineers jokingly mentioned, to a client, how they turned a ‘cranky Augie’ to a ‘happy Augie’. The client fell off his chair though not for the humour of my engineers, they are not that funny. He was serious. He wanted to see the platform. I showed it on my phone and there was our first client for this “hiring platform” for which we didn’t even have a name then. Not even a code name.
So, we called this ‘hiring platform’ Happy Augie. (for obvious reasons)
Result of the Revamped Recruitment Process
The dashboard makes our recruitment process much more efficient without compromising the overall quality of the candidates chosen for personal interviews.
- Candidates can apply to jobs directly on our website after seeing an job posting on job portals
- Applicants for different positions are listed in a tabular form. Relevant information from their resumes are displayed on the table using a resume parser powered by Natural Language Processing.
- Applicants are ranked based on the job description making the process of finding the best match simpler.
- The platform automates interview invites, interview schedules, updates and feedback. Candidates can re-schedule their own interviews without calling or emailing.
- There’s a scope of pre-assessment through one-way interviews where candidates record their videos to customisable pre-set questions. This saves our time and the candidate’s reducing the number of personal interviews.
It turned a cranky me to a happy me, “Happy Augie”.