Understanding AI in recruitment

What exactly do we mean when we talk about artificial intelligence (AI)? Here’s a simple definition from Mona Khalil, former Senior Manager, Data Science, at Greenhouse: “AI is an automated system that performs a task you’d typically expect some human intelligence to perform.” And while there’s been a lot of buzz in recent years about generative AI like ChatGPT, other forms of AI like machine learning have already been around for quite some time. In fact, if you’ve ever seen recommendations from news sites or streaming services, those are examples of machine learning in action.

When it comes to the role of AI in recruitment specifically, there are a lot of conflicting feelings. A report from Harvard Business Review shows that 91% of key HR decision-makers “believe optimising hiring processes with automation and AI is necessary for long-term business success”. At the same time, research from Pew on AI sentiment in the US shows many people reporting feeling “wary and uncertain of AI being used in hiring and assessing workers”.

Greenhouse data reflects these mixed feelings. In a survey of candidates and HR leaders, we found that 62% of HR leaders believe that AI can help them hire the best candidate. At the same time, 31% of candidates are worried that a company using AI might reject their application.

Here’s how Madeline Laurano, Founder of Aptitude Research, sums up her take on the role of AI in recruiting: “The more that you can become a champion for AI, get curious about AI, question AI, become sceptical about AI, the better you’re going to be at your job … and [improve] the candidate experience as well.”

‍

Challenges and risks in AI recruitment

While AI offers speed and efficiency, it also comes with challenges, risks and ethical considerations. There is still a lack of trust in AI, especially when it comes to hiring decisions. 

One of the biggest concerns among candidates and talent acquisition professionals alike is the potential biases that may influence AI recruitment tools. “These models are only as good as the data they’re based on,” said Tony Hobley, former Chief DE&I Officer at Omnicom Precision Marketing Group.

What impact might this have on hiring? Here’s how Dr Stacie CC Graham, the former Global Director of the Racial Equity Programme at WPP, explained it: “If we look at what made someone successful 20 years ago, that will have excluded a lot of people who may have been caretakers, who may have had physical conditions that kept them from being able to be in the office 10–12 hours a day. That’s already in the data, so if machine learning is based on what made someone successful in the past, it’s likely going to be based on things that many of us would advocate against today.”

And when it comes to generative AI like ChatGPT, it’s worth considering that it tends to create content without a lot of variability or individual expression. Mona Khalil said, “If you just take the recommendations of a generative AI system like ChatGPT at face value, the content out there is going to look the same.” This can make it harder for both jobseekers and companies to differentiate themselves if they’re both relying on generative AI for their written communication.

What can you do to mitigate these risks and ensure fairness with AI recruitment? Make sure that any tool you use has comprehensive bias monitoring. You can – and should – also actively monitor the impact of your use of that tool. 

‍

AI tools for recruitment

AI tools can enhance efficiency in almost every step of the recruiting process. Here are just a few examples of popular AI-powered recruiting tools:

• Sourcing and surfacing candidates

AI sourcing tools can help talent acquisition professionals source candidates who have the skills and background they’re looking for. Instead of conducting lengthy searches themselves, they can simply provide a job description or list of desired skills to the tool and it will return relevant results in a fraction of the time. Some of these tools can also automate outreach to kick off the conversation with candidates who appear to be a good match. There are also tools that use AI to connect active jobseekers with hiring teams.

• Filtering candidates

You can filter candidates through smart searches (available within Greenhouse Recruiting). This makes it easier to narrow down candidates in a transparent, explainable and compliant way using objective information like the suggested keywords based on the public job descriptions. Ariana Moon, Vice President of Talent Planning & Acquisition at Greenhouse, said, “This is a game-changer because there are so many inbound applications right now.” 

• Anonymising CVs resumes (to limit bias during screening) 

CV anonymisation is an AI-based recruiting tool that redacts personally identifiable information, such as first and last name, gender and candidate photo. Using a tool with this feature reduces the likelihood that you’ll be influenced by unconscious bias and gently nudges you to focus on candidates’ skills instead.

• Generating interview insights and summaries

A tool like BrightHire’s Interview Assistant provides interview summaries and highlights, which eases some of the burden of taking notes during interviews, making it easier for hiring managers and other interviewers to be present during their time with candidates while still making data-backed hiring decisions.

• Writing more inclusive job descriptions and employer branding content

Tools like Textio use AI to identify biased language with a large language model. You can run any written text through this tool to identify biased or problematic language and get ideas on how to reword it.

“While AI recruitment software has undoubtedly revolutionised the hiring landscape, it’s essential to remember that even the most advanced AI solutions can’t guarantee a flawless interview experience,” wrote HR consultant Steve Goldberg. In other words, no AI tool can be a substitute for a thoughtful and well-structured recruiting process.  

‍

The role of AI in talent acquisition

Now that we have a better understanding of some of the AI tools that are out there, let’s consider how AI is impacting talent acquisition in general. 

One of the most important things to understand about AI in talent acquisition is that it’s not just TA teams that are turning to these tools – candidates are increasingly using AI, too. And this is transforming the TA landscape.

AI has made it easier than ever for candidates to apply for jobs, with 38% of jobseekers mass applying to roles, flooding employers with CVs rather than pursuing targeted opportunities. According to our own research at Greenhouse, 35% of candidates said it felt fair to use AI in their applications since companies were probably using AI to sort through their CVs. Fair or not, the end result is the same: driving up the number of applications for every open role and increasing recruiters’ workload.

The good news is that there are plenty of ways AI can save time for TA professionals, including by generating first drafts of job descriptions and candidate outreach, summarising the key points from interviews, and removing the back-and-forth of interview scheduling by automatically suggesting optimal meeting times.

If you’ve been wondering how you might start using AI for talent acquisition (if you aren’t already), be aware that you may encounter resistance from both a technical and an ethical perspective. When it comes to the technology, Mendy Slaton, People and Talent Operations Leader at Lattice, said that at Lattice they’ve been creating short video tutorials about use cases for AI and sharing them widely. This has helped people feel more comfortable understanding what AI is capable of and how to use it.

If the resistance you encounter is more along ethical lines, you might find it useful to start a cross-functional AI committee that can identify safe ways to experiment with AI in your company. 

‍

Balancing AI and human-centric recruitment

At its core, recruiting is about finding the people who will help your company succeed – it’s a human-centric process. That’s why it’s critical to find the right balance between AI and human involvement. 

Here at Greenhouse, we’ve given a lot of thought to this topic and published guidelines for using AI in our interview process. Here are a few key points: 

Our Talent Acquisition team takes great care to be methodical with how we use AI in hiring. We have no evidence to make us believe that AI is capable of making reliable end-to-end hiring decisions without human intervention, and therefore we believe AI should supplement – not replace – human judgment and decision-making. As AI comes with flaws and risks, including bias, we view it as a co-pilot that can make hiring teams more efficient, not an autopilot that eliminates the need for human oversight.

We also outline some of the ways we use AI along with human TA professionals to boost efficiency and create a positive candidate experience. These include:

• For job descriptions, interview questions and sourcing messages

Our hiring managers and recruiters may use AI-generated content as a starting point, but our final product does not reflect an unaltered AI-produced output – it always incorporates critical human input.

• For CV parsing

Greenhouse uses AI in our CV parsing process to locate CV details that may identify a candidate. This is used to support our CV anonymisation functionality. This means some CV details, like names, are kept anonymous to create a fair and equitable evaluation of candidates.

• For interview transcription and summary

With candidate and interviewer consent, we use AI to transcribe and summarise our interviews, which lessens the reliance on the note-taking skills of an interviewer and mitigates the hiring delays and potential bias caused by incomplete documentation of interview feedback.

‍

Providing guidance to candidates on their use of AI

Finally, we believe that this transparency should go both ways. We share the ways our TA team may use AI in the recruitment process, and we also let candidates know our expectations about how they can use AI throughout the application and interview process. Our guiding principles here are: 

• We want to know the authentic – not artificial – you
• Two-way accountability
• Prepare vs perform

As Ariana Moon, VP of Talent Planning & Acquisition at Greenhouse, summed it up, summed it up: “We design our interview processes to get to know the candidate’s own intelligence, skills and experience, because we’re most interested in their independent ability to think and perform in real-world settings and how that applies to the job we’re trying to hire them for.” 

‍

The future of AI in recruitment

How will AI continue to shape the future of recruitment? Former Greenhouse Chief Product Officer Henry Tsai outlined how we’ve been considering this topic at Greenhouse and how we view the emerging trends and advancements in AI for recruiting. Here are a few highlights from what he shared.

• AI content generation

Large language models (LLMs) already enable recruiters to go from zero to one in the structured hiring process by creating job descriptions and interview questions. In the near future, we envision our hiring software will have advanced text-generation models built directly into our platform to power faster and more effective hiring.

A recruiter who wants to generate attributes for a specific role can use generative AI to select the correct prompts to create this instantaneously while reducing human bias at this stage of the process. The quality of job postings and candidate outreach can also improve with generative AI because it can tailor outreach to each unique role and use a consistent employer brand voice.

• Categorisation: reflecting human intention

The biggest challenge with traditional CV parsing (categorising specific fields from different CVs) is keeping that process fair since everyone’s CV is different in terms of formatting and language choice.

Applying generative AI data analysis now makes it possible to gather the intention behind CV terminology. AI-powered candidate search tools can help recruiters identify potential candidates with related skills and past candidates who reached late stages in similar positions. Grouping related categories – something that used to be challenging for a machine to do without individual instruction – is now much easier. This unlocks equity among all the CVs coming in for a specific role. It also makes skills-based hiring much easier since recruiters can source from a broader set of relevant candidate experiences.

• Summarisation: using AI to transform data into insights

Giving our users the ability to have a natural language conversation within our system means saving hours and hours of a recruiter’s time. Imagine typing in a sentence to find the best technical candidate who’s also gotten the most “strong yes” scorecards from the team. Or asking for a summary of reports so you can more easily track your hiring efficiency over the last three months. Using AI, our customers can now take a fresh look at all aspects of the structured hiring process in a fraction of the time.

Another exciting area in automation with AI is summarising interview transcripts. Think about the potential for recruiters being more present in their interviews, knowing AI will be there to help them with note-taking and synthesising conversations with candidates. Having an objective view of what someone said in an interview will also help reduce bias. This benefit will be felt by candidates, too, with more focused recruiters able to put less time into transcribing answers and more energy into getting to know each person on a human level.

• Automation: streamlining complex tasks

If you’ve ever tried to schedule a meeting with more than one other person, you’ve felt the pain that many recruiters feel when setting up times to meet with candidates and hiring managers.

Imagine if the first conversations around panel scheduling were done by the prompting engine in a natural language conversation: “Find my team a time to have a panel discussion about a candidate that suits everyone’s time zone, prioritises free time over busy and doesn’t book over lunch.” This type of system would break down barriers and interpret calendars more efficiently.

Another use case involves automating the applicant flow at the top of your hiring funnel. This system could turn off the inflow of applications – and more crucially, turn it on when you’re not getting enough applicants or start buying job ads programmatically.  

‍

Preparing for the evolving role of AI in recruitment

A quick recap of the Greenhouse perspective: right now, we don’t believe that AI is capable of making end-to-end hiring decisions without human interventions. There’s just no good business or moral reason to hand the wheel to AI when we are aware of its existing flaws and risks. That’s why we’re intentionally investing in research that drives ethical and sustainable hiring, where AI can assist, but not replace, hiring decisions made by human beings. 

No doubt, the advancements in AI technology are transforming hiring at every level. It’s important for any company in the hiring space to use AI as a way to make things better, faster, fairer and more efficient – and to do so in ways that benefit companies and candidates alike. At Greenhouse, we’re ready for what’s next. We’re committed to helping our customers utilise advances in AI to get better at hiring, while staying ethically responsible and sustainable.

‍

‍

‍

‍

Read more featured resources here:

• A buyer’s guide to applicant tracking systems
• Buying an ATS: Everything you need to know
• Seamless integrations and tech stack guide

‍