LinkedIn will deploy AI agents to connect recruiters and potential candidates on its platform.
Hiring Assistant, LinkedIn’s recruiter agent, will read job descriptions or written prompts from recruiters and hiring managers and then suggest candidates based on specific criteria.
Hari Srinivasan, vice president of product for LinkedIn Talent Solutions, told VentureBeat that recruiters often spend so much time writing emails and messages to potential candidates and copy-pasting job descriptions on different platforms. He said this type of work keeps recruiters from doing the most meaningful part of their job: recruiting new employees.
So, when LinkedIn began building Hiring Assistant, Srinivasan said one of the goals was to make it easier for recruiters to find talent that fits their requirements instead of making a lot of preparations to reach that talent.
“What’s important is that these are not just recommended matches, it needs to actually go through and start to evaluate each of these profiles,” Srinivasan said. “It’s summarizing the candidates and saying if this person is a good fit or not based on their qualifications.”
LinkedIn’s focus on AI combines growing trends in the hiring space. Companies like Micro1 have released AI-powered hiring and interviewing platforms to streamline the hiring process. AI agents have become a big trend for many enterprises, and there seems to be no stopping its growth.
Orchestration layer of recruiting agents
To do this, LinkedIn deployed AI agents. Recruiters will write a prompt like “I’m looking for an engineer with experience in machine learning and product management at scale” or bring in an existing job description. An agent will read the prompt and other recruiter notes and translate these into role qualifications. The agent then builds a pipeline of candidates, even identifying previous applicants.
Erran Berger, vice president of product engineering whose team built Hiring Assistant, said LinkedIn had to embrace that AI agents are non-deterministic and that humans need to be in the loop. His team also had to figure out a way to create an orchestration layer so the agents could use their reasoning capabilities to take tasks and break them down.
One way they figured this out is to build experiential memory; basically, the agent’s model remembers previous interactions with the recruiter and adjusts how it looks for candidates based on this feedback. Berger said eventually, the agents learn different preferences for open roles. It also means there would be many subagents for each job opening.
“Right now, the workflow is pretty straightforward, but as we develop more and more capabilities, it’s not gonna look like a simple straight line,” Berger said. “That’s why we built a meta-agent capability.”
LinkedIn has been leveraging generative AI for some time now. Last year, it unveiled AI chat tools that let users use AI to generate messages, profiles and job descriptions. Reid Hoffman, the company’s founder, also recently spoke about his concept of “super agency,” where AI is more of a tool for humans than a replacement.