AI in development recruitment: Time-saver or barrier to inclusion?

Artificial intelligence is streamlining development recruitment, but it’s not a solution for fairness or fit — and in some cases, hiring organizations tell Devex, it may even be getting in the way. “Its rapid development has raised both opportunities and challenges, especially in hiring. The public accessibility of AI tools has, at times, made it harder to assess candidates fairly,” a World Health Organization spokesperson told Devex in an email. The likes of AI programs such as ChatGPT, Lindy, and HireVue are being used to write job descriptions, screen resumes, source candidates, create assessments, and conduct interviews. This is particularly helpful in the development sector, where many organizations might be based in low-resource settings or have a scarcity of human resources, said Caitlin van der Merwe, a talent and diversity consultant who has helped organizations such as BroadReach and Vantage Health Technologies with recruitment. It can also help when hiring needs to happen quickly to deploy staff in an emergency context. “Here, AI is not only appropriate, but potentially life-saving as it accelerates operations where speed is mission-critical,” the WHO spokesperson said. But all interviewees stressed that while AI can streamline the administrative load of hiring, if over-relied upon, it can lead to poor hiring decisions that exacerbate biases and create a barrier to localization. “Any NGO that utilizes AI extensively in its recruiting processes will have to double down on continuous monitoring of outcomes and metrics analysis to ensure no disparate impact occurs due to the utilization of AI in recruiting,” said Bo Young Lee, an AI and workplace ethics expert. “AI can’t replace the human intuition needed to assess cultural fit, leadership potential, or the deeper motivations behind candidates’ career choices.” --— Dela Atubra, managing director, NGO Recruit The pitfalls of AI in development recruitment AI should never replace a human when it comes to the task of comparing candidates, said Katja Schramm, director of Mission Talent, which specializes in executive hiring in development. AI recruitment tools — such as Recruit CRM, Workable Recruiting, and Greenhouse — have been developed in Western contexts and therefore contain Western biases that may exclude candidates from elsewhere, said Selin Nugent, assistant director at the Centre for AI, Culture and Society at Oxford Brookes University. For example, credentials and experience gained in different contexts may not be equally recognized, manual applications might not be weighted the same as those submitted digitally, and some programs may eliminate applicants based on lower language or digital capabilities. While for certain roles, such as a monitoring and evaluation officer or grant writer, said skills might be important, for the likes of a local program officer, they might not be. According to the African Union Development Agency, only half of African countries include computer skills in their school curriculum compared to a global average of 85%, which means that when AI assesses an application from a person in Africa for a local position versus someone from the global north, they may rank them unfairly. This poses a significant hurdle at a time when the sector is pushing for more programs to be led by local people. There is a push, said Nugent — who contributed to the development of the World Economic Forum’s 2021 tool kit Human-Centred Artificial Intelligence for Human Resources — for more recruitment technology to be designed within the country it is intended to operate so that it has the context of local résumés, credentials, and skills. The human element AI should also not be used to fully replace in-person interviews, said all those Devex spoke to, as it lacks the capabilities to judge character. “AI can’t replace the human intuition needed to assess cultural fit, leadership potential, or the deeper motivations behind candidates’ career choices,” Dela Atubra, managing director at Africa-focused recruitment agency NGO Recruit, told Devex in an email. “It also struggles with nuances like understanding a candidate’s lived experience or how their skills might transfer across sectors.” This is pertinent to the development sector where an individual might be required to lead through uncertainty, manage diverse teams, or navigate political sensitivities across countries. Such abilities can’t be judged by an algorithm alone,” said the WHO spokesperson. “Similarly, roles in culturally complex or sensitive environments, like leading health emergencies, humanitarian negotiations, or community-based programming, require more than matching skills on paper. They demand the ability to listen deeply, respond with empathy, and build trust across cultures. These qualities are hard to capture through data points alone.” Human interaction is also vital when looking to fill senior leadership or specialized positions, said Atubra, explaining that these roles “need a tailored, thoughtful approach that goes beyond algorithmic matching.” For Nugent, video recognition software and AI should not be used to decide whether a candidate is eligible for future roles within the same organization. It is inappropriate, said Nugent, explaining that emotions aren’t standardized across people and there are too many issues with skin color and gender. “These can play into ways that can disadvantage a candidate,” she said, “but also disadvantage an organization.” Looking ahead As the development sector looks to cut costs in times of budget limitations, the use of AI to support basic tasks is likely to increase, interviewees said. Despite this, van der Merwe doesn’t think candidates should change how they apply to a role. While tips online suggest using keywords, avoiding specific fonts, or including data points, ultimately an application should still be reaching a human, she said, meaning that the same values apply. Van der Merwe does, however, think many candidates wrongly assume all hiring is done by AI and attempt to “trick” the system. “But when they do eventually get to human contact, it's going to come out and it's not going to reflect well on them,” she said. As an example, Jeremy Schifeling, a career development speaker, author, and former recruiter for Teach for America, was recently hiring for his own company and noticed a lot of cover letters read the same — because they’d been written by AI. “My immediate instinct was these candidates are not creative, they don’t know how to use AI in a very savvy way, and they’re probably a little bit lazy; all the things that would make them completely unhireable were now on full display,” he shared, adding that for this reason he’d advise applicants against using AI themselves. For HR staff, they need to educate themselves and explore more programs beyond OpenAI’s ChatGPT, van der Merwe said. That includes being up-to-date on privacy and data utilization concerns, Lee said. “For example, Article 22 of the [EU’s] General Data Protection Regulation requires disclosure and consent for any data collected. Candidates may not want their data entered into an AI system because there are implications for how their data is used in the future.” Nugent said consortia and professional organizations should support NGOs as they explore the different types of AI technology, laying out what might be beneficial and the best ways in which to use such tools. “You have to be a defensive buyer and think about what risk it is imposing on your organization, what risk it's imposing on the people that are working with your organization, and your reputation,” she said. Jessica Shatzel, senior director and head of talent management at the Orr Group, which recruits for nonprofits, said she would recommend all organizations create an AI policy before they embark on their AI journey “to make sure that they're doing this in an ethical and consistent manner.” The Orr Group only implements a tool at scale if it’s been rigorously tested and the team has agreed that it falls within the organization’s AI usage policy. “The best way to incorporate AI into the HR function is to include humans in the loop where there is regular review, feedback … with a person dedicated to managing and reviewing the results, adjusting and making sure that it stays as up-to-date as we do as professionals,” Shatzel said. Ultimately, said Atubra, the future lies “in combining AI’s speed with recruiters’ expertise to find candidates who not only fit the role but will drive positive impact.”