The Fairest Approach to Resume Screening with AI

Introduction

Analyzing resumes with AI has garnered a negative reputation in the industry due to the misuse of black-box algorithms. These systems often produce a magic score without any explanation, leaving recruiters clueless about how the scores are calculated, rendering them useless and unfair to candidates. Talent professionals are left wondering: How can we use a parameter that we don’t understand? How can we explain that number to users from the team? Are we being fair with candidates? How can we manage the biases that AI models carry from their training data? How can we mitigate potential AI hallucination?

Brainner’s approach to resume screening with AI was created in collaboration with senior recruiters and talent acquisition managers. The premise is to use AI for tasks where it excels in productivity and impartiality while prioritizing humans in those activities where their knowledge, sensitivity, and experience are critical. This is a three-step approach where recruiters are heavily involved at the beginning for job criteria definition and calibration, and at the end of the process, analyzing the processed information and making decisions. Meanwhile, AI handles the processing of recruiters’ instructions, analyzing resumes, and providing information for decision-making.

Part 1: Job Criteria Definition and Calibration - Human Activity

After talent acquisition teams and hiring managers agree on the job description and requirements, recruiters must select the most important criteria (e.g., working experience, skills, education) that will be used for screening resumes.. The goal is to replicate the key criteria they would use to manually screen resumes. After that, they should prioritise these requirements by defining which ones are mandatory versus preferred and provide weights (e.g., high, medium, low importance)

By involving human recruiters in this step, we ensure the search targets exactly what the company needs, leveraging the recruiters’ sensitivity and expertise to deliver high-quality and well-targeted inputs to the system.

To make the process more seamless, the Brainner platform integrates with top ATS in the market and generates an initial draft of screening criteria by importing the job description already uploaded to the ATS. This draft serves as the starting point for recruiters to customize by adding, modifying, or deleting criteria, and also classifying those as preferred or mandatory.

Image 1: Criteria definition for a back-end developer role

Part 2: Resume Processing - AI Activity

Once the criteria are defined and calibrated, it’s time to use Brainner’s AI to automate the tedious task of going through each resume to check if they meet all the requirements. Instead of a magic score, this approach breaks down the problem into parts, using AI to examine each resume against each specific requirement, and determine if the candidate fully meets those criteria, along with a brief explanation.. We trained our AI models to assess if candidates meet objective criteria (e.g., years of experience in a specific role, a bachelor’s degree in a specific field, or experience in specific tech skills).

The system then displays the information in a visual dynamic table, sorting candidates by a predictable score, calculated by weighting mandatory criteria more heavily than preferred ones and considering priorities among the criteria.

By involving AI in this step, we ensure that AI is used in a discipline where it excels: extracting information, analyzing it, and making conclusions about objective matters. Additionally, we relieve recruiters of a manual task that adds no value and can be automated.

Brainner’s integrations with ATS allow this process to be not only automatic, but also in real time. As candidates apply through the ATS job posting, resumes will be imported into Brainner and the analysis will be done in real time.

Image 2: Candidate analyses, including a color coding system to indicate whether the candidate meets each criterion, justifications for each criterion, and a scoring system

Part 3: Candidate Selection for Interview Stages - Human Activity

Once the candidates have been analyzed, recruiters re-enter the process to review the output and select the best candidates. It’s important to note that in this approach, Brainner’s AI doesn’t choose the best candidates outright; instead, we provide recruiters with the tools to filter out those who do not meet the most critical requirements and focus on the top 5-10% who do. Recruiters can then easily navigate this shortlist and manually review the top applicants' resumes, deciding 10x faster which candidates to interview for the role.

By involving human recruiters in this step, we ensure that the final decision is made by humans and not by AI, and we add a final manual check of top resumes.

To offer a more streamlined experience, decisions such as advancing candidates (indicating which stage) or archiving candidates (indicating reasons) can be made directly from Brainner, individually or in bulk, and these changes will be reflected in the ATS to continue the hiring process there.

Image 3: Detailed report for each candidate

Conclusion

Brainner’s approach aims to enhance the recruiter’s job, shifting from a completely manual role that involves hours of reading resumes to a more strategic one where they spend less time on the activity and focus more on criteria definition, calibration, and data analysis. Additionally, we leverage top technologies only to replace manual tasks that can be automated and avoiding subjective decisions that are likely to be biassed. Recruiters who can adapt to this methodology, integrating new technologies into their current processes, will outperform their competition.