Can AI make the job search less grueling?

AI is everywhere on LinkedIn these days, from the company’s own hiring and search tools to the buzzy terms du jour that populate job listings and the monologues of aspiring thought leaders.

But does it prevent the sometimes isolating and disheartening experience of searching for a job? We spoke with LinkedIn’s chief product officer, Tomer Cohen, about how his team builds AI tools aimed at making those matches easier. Cohen said AI is helping to surface jobs that might not otherwise be seen in traditional search.

The Microsoft-owned company announced this week that it’s making its first AI agent, a hiring assistant for recruiters, globally available later this month, after an early-access period with select customers like SAP, Wipro, and Siemens.

Cohen also discussed the ways AI is changing the job market and how the company organizes its many agents.

This conversation has been edited for length and clarity.

There are all kinds of thoughts on what AI means for jobs right now: Anthropic CEO Dario Amodei has said AI is going to eliminate half of white collar jobs. People worry that AI is already hurting entry-level workers. How much of that are you seeing bear out in the view you have into the platform? What is the biggest impact you’re seeing?

There’s no doubt that AI is basically reshaping jobs and roles. I think “jobs” tends to be not the right lens to look at. I think “skills” is probably the best lens, because what’s happening with AI is that AI is basically automating the repetitive, low-leverage tasks that we all do, especially for knowledge work. So everything from drafting to summarizing, scheduling, searching. And we allocate roughly 30% of hours currently done within the US economy [to those tasks]. So those skills, you should expect to basically be automated. Now, if your role is mainly those skills, then that role might no longer be necessary. But for many roles, those tasks are just a part of the role. So then you ask yourself, “OK, what are the next level of skills required?” And that’s really about, like, “OK, how do I work with AI better? How do I kind of shape AI to the needs of my company?” This is where you’re seeing the notion around AI fluency, AI agency. Do I know how to bring in a lot of great AI tools and know how they all work together? And that’s where we see that new role emerging and becoming really, really powerful.

There’s the question around early career talent, and that for us is something that we’re investing [in]. So internally, we recently launched the Associate Product Builder program, which is really built around AI fluency and agency, to basically invest in early career talent. But you can see that kind of broad spectrum around, “Do you have a growth mindset? Do you have AI fluency? Do you have AI agency? Are you able to be fluid with the changes of your role?” all the way to, “How do we invest in early career talent to be able to kind of groom them into those roles and that pace of change?”

Do you have ways of measuring jobs that are getting surfaced by AI tools that wouldn’t have otherwise been widely seen?

You can think of those as the long tail of jobs, in many ways. Some jobs appear in all the searches, but it’s mostly because people don’t share a lot. They’re like, “I want a marketing job in my area,” and there’s the ones that most people apply to. But if I know more about your skill set, your aspirations, there might be a job that actually not many have applied to, but is a great opportunity for you. So instead of a lot of job seekers applying to a few jobs, which, by definition, will create dissatisfaction, you’re able to actually spread out the supply and demand, so there’s a better satisfaction on both sides of the marketplace.

Everybody knows that job seeking can be a slog. You have to send out a ton of applications and never hear back from most of them. And on the other side, hiring managers are flooded with irrelevant applications. What’s your ultimate vision for what the job search should look like in the future once all this AI is implemented?

When we look at the job market right now, it’s very tight. So applicants per job, if you look at two years ago versus now, they grew by 20%, so you can clearly see that there’s more applicants per job than there were before, and that gets tighter and tighter every year.

I spend a lot of time with job seekers, whether it’s in research or interacting with them—it’s a very lonely experience. It’s not rare for somebody to cry in those sessions when they share about their experiences, because it’s very emotional. They don’t even feel comfortable sharing with their partner how stressful it is.

And I think our job is really kind of taking you all the way through this journey and really making sure you feel confident you’re getting matched with the right jobs, and we really take you all the way to the job and not just provide part of the value. So anything from when you even think about something you might want, you come to our job experience and just talk about it, just say what you’re after.

Then we’ll present jobs to you, but we’ll also present what is the right match for you. It’s very deflating when you apply to a job that you have no chance of hearing back from. Or, to my point, before, the supply-demand is completely imbalanced. So really being able to provide you with jobs that might match you really well, but also the supply-demand is actually really balanced, and there’s a higher likelihood of success there.

Then it doesn’t stop there: We help you with our AI agent. We help you write the resumé well, for that job, so you can emphasize the right skills. The AI agent helps you craft a cover letter for that job. And right now, we’re actually in the process of launching an interview prep. So you can do interview prep one on one with an AI coach. The goal for us is to secure you all the way from the moment when you’re like, “I could use a new opportunity,” to securing an opportunity. And then if there’s a job that you might really want right now, but on the skill set, the job match was like, “Hey, you’re missing this type of skill. You’re missing running Google Ads.” Then we actually have a course that can help you learn Google Ads so you can take that course and feel much more confident. So before you walk into the interview, you’ve kind of sharpened all the skills you need to be able to demonstrate to your future employer.

Which models does LinkedIn use?

We use a mix of models. We work on Azure and OpenAI, but we also use Llama for some [things] and again, we use them primarily as base models. And we have our own smaller models fine-tuned, but it’s really the work we do on top that creates the value of what the member sees on LinkedIn.

As you build out more of these AI agents, how do you think about organizing them?

For us, we see it as necessarily the application you use. So if you are recruiting using our recruiting products, then basically, for you, there’s one agent that takes care of the job. Now, behind the scenes, there’s actually a lot of mini agents taking smaller tasks and doing a proper job there, but it’s completely masked for the customer, for the user; for them, there’s just one agent they’re working with.

Same on the marketing tools, same on the selling tools. When it comes to our flagship experience, the LinkedIn.com experience, it is one unified experience across the board. And the goal is to basically escort you, so whether or not you saw something great on the feed and you want to have a conversation about it, but then when you think about potentially a new job to apply to, we’ll connect the dots for you, so you can see one unified relationship you’re building.