Getting The Most From AI Fit Scores, Candidate Discovery & More
Using artificial intelligence (AI) to hire best-fit candidates faster is finally becoming mainstream in HR. But are TA teams leveraging this superpowered tech to its fullest potential?
On Talent Experience Live, Phenom Product Manager Sean O’Donnell shared tips on how recruiters can get the most out of their AI-driven tech – from using AI fit scores and discovery capabilities to training strategies that improve results over time. Watch the full episode here, or catch highlights from the show below!
Debunking AI misconceptions
First, a few words of reassurance: Artificial intelligence isn’t meant to spawn evil robots that try to take over the human race. Instead, AI and humans work together to do things like hire top talent faster.
“We design [AI] to keep the human in it,” O’Donnell said. “It’s meant to augment human processes, not replace them.”
Another common misconception? The back-end effort to get AI platforms and processes moving is just too big of an ask for time-strapped recruiters.
“The biggest thing I hear is, ‘Our ecosystem is pretty large and there’s a lot of facets, and AI is just one of them,” O’Donnell said. But AI can actually help supercharge the recruiting experience, with the potential to save tons of time once it’s up and running.
Take candidate matching, for example. By investing a few minutes to set up this criteria, recruiters can instantly generate dozens of best-fit candidate suggestions, eliminating a significant amount of time spent conducting manual searches — which means they can focus on other critical hiring activities.
Bye, bye Boolean search — Hello AI-powered candidate matching
For years, Boolean logic and search functionality was an old favorite standby for recruiters that dates back to the 1940s. Today, AI can filter and source candidates — and so much more — while also reducing the risk of introducing past biases and mistakes.
AI-powered candidate matching within the Phenom talent experience platform is based on 4 key sets of criteria: job title, skill, location, and experience. Recruiters can set preferences in each area, and the system's algorithms go to work to suggest best-fit talent it discovers in profiles and resumes.
Not only does this help TA teams illuminate the right candidate faster, it also helps future-proof sourcing and hiring in general.
To ensure it's always improving results, AI constantly applies feedback it “learns” from all of the talent interactions that occur within the platform. This includes identifying how various job titles and skills are related, and ranking skills based on how relevant they are to the job title. Through this process, AI can keep up as positions evolve to include new skills.
“A given [job] title might require a new skill that becomes much more relevant than a previous skill. These are things that the AI will adapt to and learn over time versus just staying static,” O’Donnell said.
How to train your
AI is not a set-it-and-forget-it technology. Drilling deeper into why human input is vital for a well-performing AI platform, O’Donnell shared a few tips on how you can help improve your AI over time:
Optimize job descriptions. An accurate, well-crafted job description is a must. They also serve as the original source of information to pull relevant skills into the matching criteria (the system’s job parser “reads” the description and uses NLP (natural language processing) to extract relevant details.
Think of it this way, says O’Donnell, “[AI] is getting you 80% of the way there. So that’s good – now it’s your chance to make it awesome.”
Add skills. Through conversations with hiring managers, recruiters can glean additional skills that should be included in job descriptions. By keeping them up to date, you can ensure your AI is illuminating best-fit candidates as expectations shift. And here's a quick tip: Beware of piling too many skills into matching criteria – you’ll see diminishing returns after about 20 skills, O’Donnell later pointed out. Doing so could cause the platform to identify candidates who don’t have the most critical skills for the position.
Remove skills. Recruiters should also monitor the skills AI is pulling out of job descriptions. If some of them aren’t relevant, drop them from the matching criteria. “You’re teaching the AI to be better as it moves forward,” O’Donnell said.
How are you sourcing ideal candidates?
The path to finding the right candidate isn't always clear. But there are ways recruiters and sources can help clear the fog.
O'Donnell recommends recruiters teach the system what to look for initially by uploading an ideal candidate profile – a current employee, for example, who embodies the skills and experience the manager thinks would fit best.
If a recruiter later selects talent who deviate from that profile, AI can quickly pinpoint and highlight where the gaps are.
“It can find the difference, or the delta, between the matching criteria and who you’re actually hiring," O’Donnell said. “It accounts for that change, adapts the algorithm, and gets you folks closer to who you want to hire.”
Fit scores give recruiters the speed they need
So the system is off and running, generating lists of candidates … which, depending on the position, could total in the thousands. How can recruiters quickly zero in on the best-fit candidates?
That's where fit scores come in: AI assigns a score that ranks candidates according to how well their profiles match job description criteria.
What about candidates with incomplete profiles?
This could include candidates whose resumes are in a format not easily parsed by NLP (e.g., image-heavy resumes or files scanned in as images). Recruiters need a heads-up to keep a lookout for these files.
The system will, however, sort on partial applies, so no leads get left behind.
Perhaps a visitor has interacted with the career site – through the chatbot, for example – but hasn’t uploaded a resume or completed an application. AI will still sort these leads using any data the system has collected on them, and suggest them as a fit if there are any matches. Then recruiters can target these leads with personalized messaging, inviting them to apply.
How AI is designed to support DE&I
Some organizations are wary of AI introducing bias into the hiring process, hurting rather than helping Diversity, Equity & Inclusion (DE&I) efforts.
But AI-driven platforms like Phenom’s are designed to prevent bias, O’Donnell clarified.
“We try to factor out any adverse impact by only doing the four non-biased reasons: skills, title, location, and experience,” he said. “Our data science team can run an adverse impact study on any client to find out, ‘Are we hitting the mark?’ There’s more than one brain making sure we stay on the leading edge of how to maintain diversity in hiring.”
For more on how to hire faster with AI, check out The Definitive Guide to Artificial Intelligence for Recruiting
Sign up to get notified about future episodes of Talent Experience Live! Catch us on LinkedIn, YouTube, Twitter, and Facebook every Thursday at noon ET to get the latest in recruiting, talent acquisition, talent management, and HR tech.