AI in Recruitment: Managing the Risks for Successful Adoption
Imagine you could enhance the effectiveness of your HR function by as much as 20%. That’s very well possible by applying generative AI in recruitment and talent acquisition, as research shows.
AI in recruitment has been one of the key applications of artificial intelligence in HR practices. Almost 3 in 4 companies use AI in their recruitment and hiring processes at least to some extent. With the rise of generative AI, this number is only set to increase.
New legislation, such as the EU AI Act and the New York AI Bias law, has drawn attention to AI’s risks to HR. Organizations must approach adopting AI solutions responsibly to reap their full benefits.
In this article, we discuss an AI Risk Framework that can be used to evaluate the trustworthiness of AI technologies, specifically focusing on applying the framework to recruitment.
Contents
The current state of AI in recruitment
Using the AI risk framework to responsibly adopt AI in recruitment
The current state of AI in recruitment
Applicant Tracking Systems (ATS) pioneered the use of AI in recruitment with a critical focus on enhancing the quality and efficiency of recruitment practices.
With the emergence of generative AI, the applications of AI within recruitment have dramatically expanded. This advanced form of AI goes beyond streamlining processes to focus on driving higher levels of personalization, advanced predictive analytical insights predicting candidate success, and a reduction of bias in hiring decisions.
Organizations such as Electrolux, Cigna, and Kuehne and Nagel have all reported significant gains in productivity, quality of output, and impact of HR by adopting generative AI.
Businesses can apply generative AI throughout various stages of the recruitment process:
Sourcing
– Writing job profiles and job adverts targeting specific audiences
– Personalizing job adverts to highlight specific organizational requirements
– Removing bias in profiles and using gender-neutral language.
Screening
– Using analytics to match candidates to job requirements
– First-line screening of candidates using interactive asynchronous interviews.
Interviews
– Preparing relevant interview questions based on job requirements and candidate profile
– Generate case studies for candidates aligned to testing particular skills
– AI-driven candidate interviews and scoring.
We discussed AI in recruitment with Qasim Asad Salam, CEO and Co-founder of Remotebase, a company that helps organizations hire remote software developers. You can watch the full episode below:
There are, however, risks associated with using generative AI in recruitment, and many cases of irresponsible use have made HR professionals reluctant to experiment. Risks related to privacy, bias, and candidate trust have led to questioning the ethics of using GenAI for recruitment purposes.
Many HR professionals report that they are unsure of how and where to apply AI, raising concerns about privacy and data security as the main barriers. Our research at AIHR indicates that many HR professionals experience a sense of inaction, needing to learn how to get started with AI or how to evaluate the safe utilization of AI systems.
Using the AI risk framework to responsibly adopt AI in recruitment
Organizations that want to adopt AI in recruitment and other areas of HR responsibly need to take a systematic approach. That way, they can address ethical and legal considerations, as well as enhance the strategic and operational effectiveness of AI initiatives within Human Resources.
The framework below, adapted from the National Institute of Standards and Technology and the World Economic Forum, highlights the essential criteria and questions that can be used to evaluate the trustworthiness of AI systems.
Using these questions, organizations can determine the level of risk associated with implementation:
- Acceptable Risk refers to instances where the risks are known and controls are in place to mitigate them. For example, there are controls in place to mitigate selection bias during sourcing reporting on demographic variables.
- Mitigation Required highlights instances where further controls must be implemented before considering adoption. For example, a process needs to be implemented to review AI decisions over a period of time aligned to critical job requirements.
- Unacceptable Risk refers to instances where, regardless of the controls, the potential for harm is significantly more than the perceived benefit. For example, where AI will cause harm due to its recommendations and predictions.
Importantly, these characteristics are interrelated, and when evaluating AI risk, they must be examined in relation to each other. Organizations will have to make trade-offs to ensure responsible adoption and effectively manage the risk of using AI as part of their recruitment practices.
Putting the model into practice
Our hypothetical company, TX Energy, is a large manufacturing business focused on developing solar power energy systems. They have grown significantly, and the workforce has grown substantially over the past three years. They are considering implementing new GenAI technologies into their recruitment process to achieve the following benefits:
- Proactive sourcing of highly critical talent, such as engineers
- Conduct interviews with high-volume recruitment applications such as customer service
- Drive predictive analytics better to understand the success criteria of potential implementation consultants.
Using the framework above, TX Energy assesses the risk and trustworthiness of the solution as follows:
Criteria TX Energy Considerations Response Risk Rating Validity and reliability During screening, does the AI system accurately screen candidates out based on specific job requirements, or are non-relevant criteria influencing screening decisions? The system has set controls that draw a sample of screened-out candidates and reports on criteria to disqualify candidates. Acceptable Risk Safety During recruitment, is AI inadvertently leading towards limiting the equal opportunity of some minority groups to gain employment? The system provides hiring managers with a dashboard that tracks talent pool demographics to determine trends. Acceptable Risk Security and resilience During record management, is the data secured with a set management policy? The system is secure and aligns with various data security protocols and frameworks Acceptable Risk Accountability and transparency What is the level of transparency of interaction and engagement with AI during the candidate experience? The system does not proactively communicate to candidates that the interactions are AI-based.
Mitigation is required in terms of communication with candidates.Mitigation Required Explainability and interpretability Are AI-based recommendations during screening and interviewing with expectations when doing quality controls? The system reports on hiring decisions, yet there is a requirement for a quality process that evaluates outputs every quarter Mitigation Required Privacy enhancements How does AI interact with the candidate, and how is personal information handled? Personal information is handled according to set data protocols, and the tone of how AI can communicate is consistently evaluated through sentiment analysis. Acceptable Risk Fairness Is AI discriminating against a particular group during the screening and selection process? Exception reporting is available to monitor hiring decisions against key criteria such as demographics. Acceptable Risk
Following this thorough analysis, TX Energy has decided to adopt the AI solution, initiating a 3-month pilot program. This preliminary phase will test the previously outlined controls, aiming to effectively mitigate the identified risks.
This strategic approach ensures that potential challenges are addressed in a controlled environment, paving the way for a smooth transition to a large-scale implementation upon successfully completing the pilot.
Final words
We are still only at the beginning of the AI revolution in HR. As organizations adopt AI more extensively in recruitment and beyond, they must have a solid foundation to evaluate the trustworthiness of AI systems and use them responsibly.
AI provides significant opportunities to HR professionals, but it’s important to balance benefits with a responsible adoption approach that considers the risks associated with these technologies and enables organizations to make informed decisions regarding their use.
Weekly update
Stay up-to-date with the latest news, trends, and resources in HR
Learn more
Related articles
Are you ready for the future of HR?
Learn modern and relevant HR skills, online
