How LLMs Are Supercharging AI-Powered Talent Intelligence

​It's no surprise that some five years after talent intelligence platforms broke onto the HR scene, AI is transforming these systems in ways that promise to revolutionize the recruitment industry. Leading companies in the field, including SeekOut and Eightfold AI, have recently announced that they have integrated new generative AI large language models (LLMs) into their systems to help both job candidates and recruiters work more efficiently.

FEATURED RESOURCE PAGE

ChatGPT and AI
in the Workplace

"Efficiency" in this world means primarily one thing: speed. Ben Eubanks, chief research officer for Lighthouse Research and Advisory and author of Artificial Intelligence for HR (Kogan Page, 2018) tells the story of a competition that pitted recruiters against an algorithm on a task of matching candidates to jobs. Contestants—including the computer—were presented with a couple thousand real resumes and three genuine job descriptions with the goal of figuring out which candidates were actually chosen. A human won the competition, but that person took 30 hours to wade through all the information. Meanwhile, the algorithm came in third after parsing the data for a mere five seconds.

The way LLMs are transforming talent intelligence platforms relies on similar logic: AI's blinding speed can make such a big difference in recruiters' workflows that perfection of output is optional.

Take for example SeekOut's SeekOut Assist, which the company calls "ChatGPT for recruiters." SeekOut maintains a database of a billion people based on publicly available data and has a very powerful search function to dig through it. Before SeekOut Assist, recruiters would need to manually write Boolean strings to conduct these searches, using their human judgment to incorporate skills, requirements and qualifications for a given job.

With the new AI-enabled tool, they can simply paste the job description into a box and click "create search" to prompt the system to automatically pull relevant job titles, build complex Boolean strings out of those titles, identify preferred skills and qualifications, and exclude job seekers without the minimum requirements.

"This isn't going to be a perfect search," said John Tippett, SeekOut's head of product, "but the idea is to go from the job description to a B+ search within 20 seconds. Then, the recruiter goes in manually to apply their expertise."

Eightfold AI's new tools—Employee Copilot and Recruiter Copilot—similarly use LLMs' ability to understand language to help recruiters and job candidates alike. Job seekers can use the tool to generate cover letters and resumes. Recruiters can input a need, such as a candidate with a certain skill or in a certain location, and have the system automatically turn it into a query.

"Instead of interacting with buttons and checkboxes, you can type in what you're looking for," said Sachit Kamat, Eightfold AI's chief product officer. "That's a pretty big advancement."

In the case of these tools and others that are changing recruiting, AI-enabled functionality is not meant to replace the expertise and sensibility of human recruiters. Their role in the hiring process is essential, not only to ensure accurate and helpful search results but also to provide something a machine never can: a human connection.

"We have the ability right now to fully automate the hiring process," Eubanks said. "But what makes a good recruiter isn't just writing search strings. A really great recruiter is always experimenting to find a deeper way to connect with someone."

Janet Clarey, principal research director for HR research and advisory services at McLean & Company, agreed. "The next logical steps for HR in using AI LLMs is understanding how to maintain human connections to help workers feel like more than just a transaction," she said.

This is especially true because most organizations that are using these types of tools are highly matrixed, large, global organizations, Clarey said. Such complex companies may find the efficiencies and cost savings of AI-led recruiting attractive. However, not only is lack of a human connection problematic for hiring, but also, leaning too heavily on computing can introduce problems like bias.

"There are some really interesting things that the technology can do," Eubanks said. "But at the end of the day, it's the leaders who have to be accountable." 

And HR leaders' best interest is served by making their teams more efficient and successful to ensure the recruiting function gets better. Allowing AI to replace human ingenuity and social sense is a sure recipe for the opposite.

"In the end, hiring is about opportunity and people's desires—and the way the recruiter and the candidate interact," Tippett said. "I don't think AI is going to change any of it. These tools hold the possibility for people to be much better at their jobs and focus on the stuff they're really good at. For me, this is about getting to a great candidate and a great outcome for a company faster."

Katherine Gustafson is a freelance writer based in Portland, Ore.