This week, AI & OJ brought together talent leaders from Deal Path, Method, Evolution IQ, Suno, CR Day, Artnet, Harvey, and a handful of other companies across fintech, gaming, construction, and creative agencies. The group ranged from a Chief People Officer at a video game studio to a recruiter at a construction startup who just moved to NYC, and the conversation covered everything from Claude workflows to why LinkedIn Recruiter doesn't work for finding architects. Here's what we talked about.
Everyone wants AI to speed up recruiting
The room split into two camps on this, and most people had a foot in both. On one side, a CPO at a video game dev company said she'd evaluated AI-enabled TA platforms and walked away from all of them. Her team still has recruiters reviewing every resume with human eyes because the tools she tested required too many steps and not enough accuracy to justify the change. She's waiting for something that genuinely reduces the work rather than just rearranging it.
On the other side, a team at Method has Claude operationalizing their entire recruiting process. They run AI voice capture during weekly hiring manager one-on-ones, and Claude automatically builds and updates job descriptions, process documentation, and candidate criteria from those conversations. It's not a single tool they bought. It's a workflow they built themselves, and it's continuously updating as hiring needs shift.
Most people landed somewhere in between. At CR Day, an AI transformation firm that just announced seed funding, the team is using Ashby's criteria to flag high-signal candidates rather than filtering out low-signal ones. They've also developed custom Claude skills for resume review and for matching candidates to the right first-round interviewer. They're using AI to surface what's worth paying attention to, not to disqualify people automatically.
A recruiter at Harvey mentioned they're scaling up AI agent work with Claude internally and building internal tools, though the specifics stayed high-level. And at Evolution IQ, the team is using Claude with the Chrome extension for LinkedIn profile searches as part of a broader recalibration. They've paused hiring to raise the bar on technical talent and are rethinking how they evaluate engineers' AI capabilities during the interview process.
The people getting the most out of AI in recruiting are building custom workflows around tools like Claude and Ashby, stitching things together themselves, and keeping human review in the loop at the stages where judgment matters most.
Sourcing beyond LinkedIn is becoming a real priority
Multiple attendees brought up Juicebox, and almost everyone who'd used it had hit LinkedIn's scraping restrictions. The tool works until it doesn't, and LinkedIn is getting more aggressive about blocking automated access. One recruiter at Suno said their usage had become limited enough that they'd moved back to traditional sourcing methods: X-ray Google searches, GitHub, Stack Overflow, and Boolean searches.
EXA came up as an alternative approach to sourcing that several people were interested in. It's a semantic search API that lets you find companies by description rather than keyword. You can search for something like "companies similar to Figma" and get a list of results based on what those companies actually do, not just what's in their metadata. Layer Fiber on top of that, and you can extract LinkedIn profile data from the company URLs that EXA returns. The workflow goes: idea of what kind of company you're targeting, to a company list, to a candidate database. It flips the typical sourcing flow from "search for people" to "search for companies, then find the people inside them."
For non-tech roles, the tools look completely different. A recruiter transitioning from IT into construction recruiting said that Greenhouse and LinkedIn Recruiter were not effective for finding construction engineers, architects, and project managers. Indeed Smart Sourcing was producing significantly better results. The salary structures and bonus systems in construction are different enough from tech that the typical recruiting stack just doesn't translate.
A recruiter at a small boutique firm focused on placing high-caliber engineers at early-stage startups described their approach to sourcing as less about finding new people and more about staying on top of previous candidates they've already connected with. They built an in-house product that functions as a stronger internal data center for tracking and re-engaging past relationships. The logic is that the best candidates for early-stage startups are often people you've already talked to, not people you're cold-sourcing for the first time.
The candidate experience conversation shifted the room
The most engaged part of the morning was about what happens after you get a candidate's attention.
A team at Method described their approach to candidate collateral. Before the first interview, they send candidates a curated packet: an article, a short deck, and a video walkthrough of how their API product works. The API isn't a tangible product in the way a consumer app is, so the materials help candidates understand what they'd actually be building. If the candidate doesn't engage with the materials, that's a data point. If they do, the first conversation starts at a higher level because you're not spending 20 minutes explaining what the company does. The key is sequencing it. You send the collateral early enough that it feels like a resource, not a test.
At Suno, a music tech company, the approach is even more hands-on. During interviews, the team creates a song with the candidate using their product and sends it to them afterward. It gives the candidate something tangible and shareable, builds a genuine connection, and demonstrates the company culture in a way that a careers page never could.
Both approaches point to the same idea: start selling from first contact, not at the offer stage. Engineers respond better to problem-first conversations than pitch-first ones. If you can show a candidate the specific technical problems the role would solve before you start evaluating them, the dynamic changes completely. You're having a conversation between two people trying to figure out if there's a fit, not an interrogation where one side holds all the information.
The room also got into Glassdoor management. One team runs internal campaigns encouraging employees to leave reviews, with the important caveat that participation has to be optional. The value isn't in inflating the rating. It's in making sure the rating reflects reality and using the feedback to actually improve internal processes. Several people pointed out that transparency about startup realities, including the hard parts, helps calibrate candidate expectations so you're not losing people at the offer stage because they didn't understand what they were signing up for.
The hardest roles to fill right now
A few specific hiring challenges kept coming up. A recruiter at Suno is trying to fill an Engineering Manager for Android with a five-day RTO requirement and equity-only compensation with no bonus. That's a brutal combination in the current market, and the conversation around it focused on how you incentivize candidates to come into the office when you can't compete on cash. The group landed on leaning hard into product mission, growth opportunity, and equity upside, but acknowledged there's no magic answer when the constraints are that tight.
A creative ad agency was building out an entirely new content creation department with niche roles that the recruiter described as "unicorn" positions. LinkedIn Recruiter wasn't surfacing the right people for these specialized roles, and they'd recently purchased new tools but were still figuring out implementation.
More broadly, the tension between thorough technical evaluation and candidate experience came up several times. Take-home projects push candidates away, especially senior ones who have multiple processes running simultaneously. But abbreviated technical screens don't always give you enough signal. One team at CR Day framed it well: the challenge is technically evaluating candidates without overwhelming them with a lengthy process, especially when you're an early-stage company competing for attention against bigger names that can offer faster loops and higher base comp.
Tools mentioned
For anyone who wants to look these up:
Claude (resume review, process automation)
Ashby (criteria-based candidate flagging, AI integrations)
EXA (semantic company search API)
Fiber (LinkedIn profile extraction from company URLs)
Juicebox (LinkedIn sourcing, with noted scraping limitations)
Gem (sourcing and CRM)
Indeed Smart Sourcing (non-tech roles)
Greenhouse (ATS)
BrightHire (interview intelligence)
Pin (multi-channel outreach with LinkedIn integration)
Fonzi AI (talent marketplace for technical roles)
AI & OJ is a monthly invite-only breakfast for HR and talent leaders, running in both NYC and SF. If you want in, RSVP here. If your company is hiring engineers, we'd love to help — fonzi.ai/schedule-call.
