Staffing Industry Spotlight: Vahan Melkonyan
Staffing Industry Spotlight is an interview series featuring leaders shaping the staffing industry, sponsored by Ascen, a leading back-office and employer-of-record for staffing agencies. This edition features Vahan Melkonyan, Founder and CEO of Jellie AI and a serial entrepreneur in the recruitment tech space.
Vahan shares his journey from building a talent marketplace in college to developing agentic AI tools that revitalize unusable CRM data. The conversation explores high-impact themes such as automated candidate marketing, the shift from manual search to relationship tracking, and Vahan’s vision for a future defined by AI-to-AI communication and high-scale, one-person staffing firms.
Francis Larson:
Vahan, thank you so much for being on the Staffing Industry Spotlight. You are the founder and CEO of Jellie AI, which is a very cool AI product in the recruitment and HR tech space. First, we want to know how you got into this space.
Vahan Melkonyan:
It’s our origin story. Together with my co-founder, we started our first company in college during our second year in 2017. It was a recruiting product, a talent marketplace to get our peers jobs. We ended up placing 2,000 fellow undergrads into their first internships or jobs. Since then, I’ve been in recruitment tech. I haven’t worked a real job in my life; I’ve just been building products in recruitment tech one after the other. Jellie is our third startup.
Francis Larson:
You started a marketplace in college. How did that go? Did the product take off? How did it lead to your next venture?
Vahan Melkonyan:
We stumbled upon the idea for that company by accident during a hackathon and decided to stick with it for years. It caught on, and at one point, we partnered with the seven major universities in our country. If someone was graduating, they already had an account on our platform provided by the university. We were selling to employers, who loved the idea of having one network to post jobs and appear to every undergrad. We worked with big brands like Coca-Cola and Philip Morris.
Francis Larson:
I believe I saw somewhere that you sold that company, started another, and then sold that one, too. Is that right?
Vahan Melkonyan:
That’s how the story goes. At that point, we were tired of marketplaces. I still don’t think I would go back to building a recruiting marketplace. We started our second company, called Cauldron, which was a pre-screening application app. Back then, we didn’t have AI as powerful as this to screen applications, but we thought the job application process with just a resume wasn’t working. We helped people create interactive applications that let candidates answer questions and share their backgrounds through an assessment process before reaching the hiring manager. It was also a way for them to earn a guaranteed interview. We built it through COVID and saw huge growth because everyone went remote and needed a way to pre-screen candidates. The market crashed a couple of times, but we brought that company back to life three times before successfully pivoting and selling it.
Francis Larson:
It’s different now because cheating is so easy with AI. Even though you can screen better, it’s an arms race of cheating. So, how did Jellie come about? What was the original thesis?
Vahan Melkonyan:
The thesis was somewhat the same, but the path there has changed. We spent eight months consulting for recruiting agencies to build custom products. ChatGPT had already come out, and people wanted us to integrate it into their CRMs. We received so many identical requests that those insights became Jellie. The insight was that the data in their CRMs was unusable. Agencies were sitting on hundreds of thousands of outdated candidate profiles that were not searchable. Even if they had been searchable, they wouldn’t have been useful because they were old.
Francis Larson:
What was the AI solution for out-of-date profiles and searchability?
Vahan Melkonyan:
The first product we built was an updating and enrichment engine. We used AI to get insights, find candidate LinkedIn profiles, and compare or merge data with outdated resumes. From there, we could dig deeper into their professional experience, understanding the companies they worked for, their industry experience, and whether they were with a company during a seed round or a Series B. We also used AI on recruiter-generated insights from their CRM, such as notes on specific roles. We turned unstructured data into structured, searchable data. Without AI, this wasn’t possible.
Francis Larson:
Today, it seems more agentic. What happened as this evolved? Why did you pick the current feature set?
Vahan Melkonyan:
It turns out people wanted more than just an updated database; they needed workflows around it. Our identity became centered on helping recruiting agencies get more revenue without adding to their plate. As a services business, the only way to scale is to work more hours or hire more people, which isn’t very scalable. Every candidate you know is good, but they aren't actively placing or following up on lost revenue. In the summer of 2024, we built a candidate marketing feature set. Recruiters designate their best candidates, and Jellie, as the AI agent, goes out to find roles across their network and the open web that fit those candidates. This generates leads, conversations, and eventually placements.
Francis Larson:
Would you reach out to the hiring manager? And how would you confirm that a candidate from five or ten years ago is actually looking to move before marketing them?
Vahan Melkonyan:
Those are technically two different AI agents. One task is to look through your database and track your relationships. When you find out someone is back in the market or already looking, you put them forward to the marketing engine. Nothing is manual with Jellie. You deploy a nurturing agent whose only job is to identify candidates who may be in the market and to follow up with relevant conversations. The agent tracks LinkedIn profile updates for tens of thousands of people without you having to do anything. It reaches out and pings you if there is something relevant. If a candidate you worked with previously just got hired for a VP role, they might now be a client. Great recruiting agencies know these patterns, and we’ve productized that know-how.
Francis Larson:
The whole world is your database now. Your agent can find out if they are ready to work. Could you close the loop entirely? How far are we from an agent handling client intake, creating the job description, recruiting, and conducting the interviews?
Vahan Melkonyan:
You absolutely can, and we are going there very quickly. I don’t think hiring managers want to talk to recruiting agents as much as they'd like to talk to their AI agents. Your recruiting agent could talk to the client’s internal recruiting agent, get the intake, and then talk to other recruiting agents who are marketing candidates. We might have candidate agents responding to sourcing agents to agree on terms. That feature is close, as candidates now have agents through auto-applier tools.
Francis Larson:
Anything low-paid in recruiting is going to be AI agents. What is on the horizon for Jellie? Where do you see this going?
Vahan Melkonyan:
We are committed to the vision of recruiting focused on these small recruiter shops that handle all of recruiting. We want to be the underlying ecosystem that allows that to happen. We will build the infrastructure to connect sourcing agents and marketing agents who meet to negotiate deals. Right now, we’re committed to building the best software for agencies to make the most money.
Francis Larson:
How do you see the different frontier lab models fitting into the product? Are you using a mix of models like Claude, Gemini, or OpenAI?
Vahan Melkonyan:
It’s a mix of models. I know some are better at logic, others at math, and others at search. Because we have many agents doing different things (some searching for roles, others merging data points), there are differences in how they approach tasks. A generation from now, with more updates, the models might become more in tune with one another, and we might use one. Right now, Claude is insane with data analysis and content, like generating emails or explanations. Gemini is perhaps a bit better at logic.
Francis Larson:
What do you think about competition and maintaining an edge in an era where it’s so easy to write software and get an MVP out?
Vahan Melkonyan:
AI has empowered competitors, but it has also empowered us. We built our sourcing agent over the New Year’s holidays. We already have customers choosing ours over established competitors. You can build software, but you have to know what you are building. You still need insight and a background to understand your customers’ problems. You can’t just tell a model to make the best candidate marketing agent; it won’t know how to do it. Recruitment has always been a red ocean market with unlimited competitors. You have to stay focused and build for your customers.
Francis Larson:
That’s great advice. Technology makes everyone more efficient. I think we’re going to see those billion-dollar one-person recruitment agencies soon. Thank you for being on the show.
Vahan Melkonyan:
Thank you.
