Staffing Industry Spotlight: Gino Rooney, Co-founder and CEO, Classet

Staffing Industry Spotlight, presented by Ascen, an all-in-one employer of record platform for staffing agencies, features Gino Rooney, Co-founder and CEO of Classet. Gino shares his journey from co-founding Blue Crew to launching Classet, a cutting-edge AI screening tool for recruiting teams. He discusses how generative AI voice bots are revolutionizing hourly workforce hiring by streamlining candidate engagement and pre-screening while ensuring a human touch for high-value interactions. Gino also explores the future of AI-driven recruiting, compliance with emerging regulations, and the evolving candidate experience.

Francis Larson (Ascen)

Gino, thank you for being on the staffing industry spotlight series. First off, we'd like to know who you are and what do you do?

Gino Rooney (Classet)

My name is Gino. I'm one of the co-founders of Classet, which is an AI screening tool that helps recruiting teams hire more efficiently.

My background has been in staffing for the past 10 years, with hiring technology. I previously co-founded Blue Crew, which was an on-demand staffing platform for hourly work. I really saw there the need to hire and build more efficient hiring processes and technology. It became increasingly difficult as we grew to continue to grow without needing to add more recruiters at the top of the funnel.

That was sort of where Classet was born: trying to use as much higher technology as you can to really have a great candidate experience while also driving efficiency in the recruiting funnel.

Francis Larson (Ascen)

So Blue Crew was a YC company (like us). I believe you guys sold it to IAC sometime in 2016 or around then. Are the same founders from Blue Crew with you at Classet?

Gino Rooney (Classet)

Yeah, so we love and hate each other enough that we wanted to work together. We've been working on Classet for three years. Four of our seven team members were long-standing employees at Blue Crew, and this full growth cycle.

So super passionate about the hiring space, took a ton of learnings. We're all remote now, but definitely have fond memories of like the early days in the small office in San Francisco.

Francis Larson (Ascen)

So Blue Crew was light industrial, maybe some gig-type staffing, pretty generalist.

What did you learn there that led to your AI approach with Classet? Was it getting applicants, retaining, interviewing…?

Gino Rooney (Classet)

I'm sure a ton of businesses struggle with this as well, but really a lot of the friction comes at the top of the funnel.

So if you're acquiring candidates through job boards or referrals, whatever those channels might be, just going from those initial 100 applicants to here are the 10 or 15 highly-engaged, highly-qualified folks is where we saw a lot of bandwidth going towards, especially as we were in the early days we'd hire junior recruiters to help with that process. It would continue to take up a lot of time. So you mentioned texts and mobile. We always at Blue Crew were looking for opportunities to leverage this kind of technology at that front end of the funnel. Very quickly early on actually the first iteration of Blue Crew was basically just like a texting engine to find jobs. You could sign up and Blue Crew would text you: “Hey you want to work this job this shift?” If you said yes you'd be put on the job, we’d send the directions.

Then we very quickly moved to a mobile app and got adoption with that too. Having a super mobile-first hiring process–this sounds obvious–but it's paramount to any sort of on-demand hourly solution. How you're engaging the candidates both automatically and with your recruiting team is an important piece of that too to make sure that you're funnel is consistently moving forward quickly.

Where AI screening was born from this was over the past three years, there are a lot of really cool tools out there with generative AI that help you not only engage candidates automatically in answering questions, because candidates have a ton of questions that are searching for jobs, but also to screen them and ask them questions about their resume and things like that.

You mentioned the hourly workspace. One thing that's pretty common in it is that folks don’t necessarily have the best resumes, or they don't have resume training, or they don't have a resume at all.

So this is where we think AI screening can actually really help candidates: helping them display their skills better and communicate their background, their work experience and where roles.

Francis Larson (Ascen)

If I understand correctly, a big problem in the hourly space is that there are many applicants for the job, and they need to be narrowed down to the few who would actually be good at it. Since this population doesn’t necessarily have resumes, recruiters would need to ask people questions to find the right candidates. But now, with Classet, the AI solution is taking that same model and using a generative AI to ask similar questions and respond to the worker.

Gino Rooney (Classet)

Yeah so more or less, when a candidate applies, our solution will immediately reach out to them saying: “Hey, thanks so much for applying for the warehouse associate role. We use an AI-based recruiter to help speed up the hiring process. Are you free for a quick five minutes to talk about your background?” if the candidate doesn't respond to that, we'll continue to engage them and send them reminders like what a human recruiter would do. When they do become engaged, or if they say yes right away or call the number, we'll have a script that our employers can customize, and then we'll walk the candidate through a series of questions.

If they have a resume, it'll go through the candidate's resume with them, and then establish the next steps with them.

So yeah, to your point, it's very similar to what in the early, early days of Blue Crew, what we individually were doing with our candidates, but eventually scaled up as well.

Francis Larson (Ascen)

The generative AI approach used in voice bots is interesting. Unlike past chatbots that relied on rules or predetermined responses, this AI approach is much more sophisticated.

I tried out the Classet interview bot at Staffing World, and it was super impressive.

It raises the question of how to handle candidate questions during AI interviews. What range of candidate responses can it process?

Gino Rooney (Classet)

So it's very customizable o answer your first question about like how does it answer questions? So The AI recruiter is pulling in data from your job description There are obviously requirements with that, like we want to have like the wage so we can answer questions for candidates because that's very important if you have broader questions that go beyond the job. You can upload your FAQs: whether it's company-wide or job-wide.

So, when candidates ask questions, we can retrieve that information quickly and provide the right answer. This is really important for the candidate experience.

When candidates are applying to 100 jobs online and hop onto a phone call with a recruiter, they want their questions answered before they're going to answer your questions.

So being able to very flexibly answer those questions is really important. And then once the candidate is ready to move forward to learn more about the job and tell them about themselves, that's when you have a highly kind of customizable script that our employers will have their own questions that they wanna ask or go through the work history and things like that.

Francis Larson (Ascen)

It's such a good use case for generative AI because you can give the LLM documents and have it answer questions about those documents, which is perfect for this case, right?

Gino Rooney (Classet)

Yeah, and I think an important point you made about just what's different this go around versus chatbots and the decision tree [approach].

We're really excited about just how good voice AI has become. We see this a lot in the hourly hiring space, but candidates get text fatigue quickly.

You won't get that much information from the candidate by having them answer 15 questions over text.

Rather than answering a bunch of questions over text, it's hop on a quick call, talk about your background for five minutes. You get so much more rich information out of the candidate, they're also able to communicate their skill set a lot better and quicker.

Francis Larson (Ascen)

That's so interesting. The text approach was never the full solution–it seems like voice is the future of this. Voice AI is extremely good at answering questions from documents, but how do you deal with unexpected, off-script questions?

Gino Rooney (Classet)

Yeah, and it's an awesome question. I mean, the nature of generative AI is that it's generative. Putting guardrails in place to stop these things or course correct when they do happen is really important.

It's been trial and error, figuring it out as we go. We've done a couple of things to help with that.

So the first is we do our own simulation testing. We have our own voice recruiter candidates who are calling into our voice recruiter to try to derail the conversations and get responses that aren't the right response, and kind of prying on that. And then we do do human in the loop sampling as well. We've got folks on our team who will sample the calls, listen to them, and ensure that nothing unruly is happening.

But a lot of it is, by nature of it being generative AI, just a lot of iteration and trying to put the right guardrails in place and then doing a lot of monitoring. We're also going through some audit processes as well right now with a third party to look into AI bias audit because that's important too.

Francis Larson (Ascen)

I want to talk about that in a bit. But want to touch more towards this question where if AI doesn't have the answer, how do you deal with that?

Gino Rooney (Classet)

Yeah, that part, I didn't answer fully. So, if we don't have the answer, Joy's [the AI Recruiter’s name] actually very good and has the instructions to not make up answers.

In those cases, the candidate will be told something like, "You know, I don't have that information handy. Let me take it out, and we'll get back to you."

Francis Larson (Ascen)

Okay, well, then they take it offline, and you can have the actual team follow up.

Gino Rooney (Classet)

That's where I think there's actually some power, over the long term, in your ability to train your recruiter to become as good as your best recruiter is because you have all of these calls and their recordings and the questions that are happening, you can constantly be making that recruiter better by adding FAQs that are coming up on those calls so that you can have a candidate getting every question answered that they want about that role before deciding to move forward.

Francis Larson (Ascen)

Would you call that prompt engineering?

Gino Rooney (Classet)

Yeah, I guess you could say that. In some ways, it's just providing the model with better data rather than actually changing the prompt.

But more or less, it's just us trying to continue to improve an employer's recruiter by feeding it with more information for the candidates.

Francis Larson (Ascen)

But it's really interesting. One day maybe you'll have the perfect recruiter because they'll know all the frequently asked questions and all your company data.

Gino Rooney (Classet)

Maybe the perfect recruiter is very different for each company.

I think that's important. Gathering feedback from the employers as they're building out, the recruiter, is important, but it's going to be so customizable in the next few years.

You're going to be able to, if you want to, use your own team's voice as your recruiter. You can customize all the languages and things like that. But I do think you're right that like over time, every team is going to have their version of what's considered like their own perfect recruiter.

Francis Larson (Ascen)

Where do you think the [LLM] models struggle right now? And where do you think, like, do you expect of the next models to be able to overcome that?

Gino Rooney (Classet)

I don't know if I've blamed the models necessarily because to be honest, they're, they're really good. Like they, outside of, you know, over time, having a few hallucinations that you need to improve upon, which I'm assuming are going to continue to get better, the costs keep going lower, which is really cool to see as well, which gives opportunities to pass those costs on to your customers too.

I think, especially in the voice and conversational AI space, where there's the biggest room for improvement right now, and it is improving, is making the conversation more natural.

So that's not necessarily the model's output. It's like when the model decides to interrupt the candidate, like the turn-taking kind of capability, what the latency is. Because a lot of the models have decent latency and seeing that continue to improve sort of a little bit of an assumption, but I think it's just like how you make these conversational AI's more natural sounding, and a lot of that is like how interruptive it is.

How good is it at knowing when to take its turn talking? That's where I see the biggest opportunity for improvement because if it doesn't do that, then it just kind of sounds like a voice tone that's taking you through a menu.

So having very natural-sounding turn-taking is really important.

Francis Larson (Ascen)

So you don't see issues where the AI doesn’t understand the candidate due to audio quality or language issues?

Gino Rooney (Classet)

Yeah, the the speech-to-text and text-to-speech rotation is pretty solid. We've got tons of examples on our platform where a candidate starts a call saying hey my english isn't great.

We do this in Spanish, and it'll fluidly move into Spanish and talk to the candidate that way.The speech-to-text and text-to-speech is good, rally good. There are lots of players in the space working on it. I assume that it will continue to get better.

Francis Larson (Ascen)

Where do you go next for Classet and recruiting in general. Is it something on the staffing client side like relationship management? Where do you think the AI is going to go next?

Gino Rooney (Classet)

I haven't touched on this too much to this point, but I do think it's an important thing to say that I do, and we do believe in the value of recruiters and their importance in the hiring process.

But I do think that long term, there is the need for recruiters for what we call the last mile logistics of a candidate who has done their initial pre-screening, looks qualified, and is excited about the role.

That's a great point for a human to get involved and help build a relationship, get them more excited, get them on site for an interview, and help them through that way. I do think there is still a need for human involvement. I don't think we'll probably go so far down upstream in the hiring process that we have an AI recruiter that's doing, a lot of the cultural, important high value touches that a human can do.

So, it is kind of we view it as a symbiotic relationship. I think there's a bigger opportunity to continue improving things, such as using automation to improve the recruiters' workflow within applicant tracking systems and their current systems. We’ll always be anchored in improving the candidate experience.

I think it'd be foolish to say that, with all this AI recruiting stuff happening, the candidates' job search experience isn't going to change over the next five to ten years.

Moving downstream into, like, onboarding the workers, or compliance, it’s too soon to say, but I think where there's a lot of friction right now is in the job search experience and this is where AI generative AI can be super helpful.

Francis Larson (Ascen)

You're probably well aware there are several states starting to regular AI for hiring. How do you see that developing?

Gino Rooney (Classet)

I think the one that we have our eye on is the New York City law around automated employment decision tools. We take kind of a two-pronged philosophy to this. The first is our product right now does not do any like sentimental analysis or make statements about the candidates fit for the job.

We do this for a couple of reasons. One is that's an area where AI bias can really creep in. If you're giving AI tools the ability to opine on how good this candidate in a subject way. So all of the evaluation criteria on the call is based on binary things that were gathered from the call, like does this person have a driver's license? Are they over 18 years old? So we're not providing an A rating for professionalism or a B rating for skill set fit.

The other reason we're doing that is that I feel like a lot of these tools give you these rankings that aren't very great or helpful. They might be directionally correct, but seeing that a candidate is a 92 versus a 95 in an ATS. It doesn’t really matter.

The way we think about it is just doing very binary evaluations. In terms of where the AI world is heading in terms of regulation, I do think it's important because hiring is such an important space for people's livelihoods. It makes sense to have some form of regulation around what AI can and can’t do, especially as it comes to making a decision not to move forward with the candidate.

So I think it'll probably continue to trend in the direction of how some of these New York City rules are shaking out, which at its core is essentially saying that if you have an automated employment decision-making tool, you should have it audited by a third party and show that there's no disparate impact and that it holds up in terms of not implementing any bias into that hiring system.

Francis Larson (Ascen)

It sounds like your approach is more of a pass-fail binary on more objective things, which sounds really good.

Gino, thanks for being on the show. It was all really great content, and we hope to talk to you again soon.

Gino Rooney (Classet)

Thanks Francis. It's great to chat.