Victor Luo conducted 800 interviews with candidates sourced by recruitment agencies during his years building machine learning teams at TikTok. By his own accounting, the experience was expensive, occasionally fruitful, frequently maddening. Now, with two co-founders and a spot in Y Combinator's Winter 2026 batch, he's built what might be his revenge: an AI agent that promises to do the recruiter's job at half the price.
They named it Paul.
Perfectly—the three-person startup behind Paul—describes their creation as "the first autonomous AI recruiter." It sources candidates, drafts personalized outreach, conducts initial screens, manages engagement through the hiring funnel. Interview-ready prospects, the company claims, simply materialize in your Slack channel. The metrics Perfectly advertises are the kind that make traditional agencies wince: 10× candidate volume, 2× higher interview pass rates, 20× recruiting efficiency (compressing twenty hours of work into one). Twenty hours of work compressed into one, if you believe the pitch.
Whether you should believe it is another matter entirely.
Undercutting the Old Guard
Perfectly calls itself an "AI-native recruiting agency," which means it still operates on the familiar success-fee model—you pay only for completed hires. But the pricing structure represents a direct assault on incumbents: 15 to 25 percent per placement depending on role difficulty, roughly half the industry standard. For fellow Y Combinator companies, the terms got even friendlier. Around January, an offer circulated on Launch YC: undercut your existing agency by 50 percent on the first role, or lock in an exclusive rate going forward.
The customer list, as of mid-April, skews heavily toward the YC universe. Corgi, Giga, LlamaIndex, Mintlify, Porter—mostly startups that know the founders or run in adjacent circles. That's standard for an early-stage company building initial traction. What's less standard is the operational claim: that an AI agent can genuinely replace the human judgment, relationship-building, and pattern recognition that recruiters insist differentiates the good ones from the spam artists.
Paul handles sourcing across platforms, writes what Perfectly describes as "personalized" outreach (though the degree of actual personalization remains unclear), screens candidates through automated conversations, and keeps engagement warm until interview day. The result, per company figures: 4× faster time-to-hire and a 250 percent lift in candidate interview retention. The company's website features a testimonial attributed to "Caleb from Series A Stealth Startup," mentioning abandoning Paraform and Juicebox in favor of "velocity from Perfectly," though like most early-stage startup testimonials, it's self-reported and undated.
Velocity is one thing. Quality is another, and the metrics—impressive as they sound—lack third-party validation at this stage. That's typical for a startup barely out of the batch, but recruiting remains notoriously hard to benchmark. What one company considers "interview-ready" another might dismiss as a waste of calendar time.
The Founders Know Their Way Around ML Orgs
The Perfectly team brings machine learning credentials from TikTok and Meta, which perhaps explains their comfort betting on autonomous agents to navigate something as human-centric as talent acquisition. Zhuang (Gary) Luo worked as a machine learning engineer at Meta following a TikTok internship. Huimin Xie served as tech lead and senior MLE at TikTok. Victor Luo, the CEO, previously built ML organizations at TikTok—hence those 800 agency-sourced interviews, an experience that likely shaped the team's thinking about what works and what doesn't in recruitment.
Their YC partner is Tyler Bosmeny, and Victor's LinkedIn activity during the batch followed the standard founder playbook: recapping the YC interview experience, chronicling the co-founder search, promoting the exclusive agency discount. By March, the team had expanded their ambitions beyond a single AI agent.
Enter Parker: The Candidate-Side Agent
Around mid-March, Perfectly unveiled Parker, a second AI agent that flips the model entirely. Where Paul serves employers, Parker serves job seekers—functioning as what the company calls a "career super-connector" via iMessage and WhatsApp. Candidates submit an intake form with phone number, LinkedIn, resume, job preferences. Parker then drafts outreach messages and attempts to create warm introductions to hiring managers on the candidate's behalf.
The dual-agent strategy raises eyebrows. Paul automates the employer side of recruiting; Parker automates the candidate side. It's unclear whether the two systems communicate with each other, or whether Perfectly envisions a future where AI agents negotiate placements with minimal human oversight. The setup suggests the company is building a two-sided marketplace where automation handles both ends of the transaction—a model that could scale rapidly if it works, or collapse under the weight of misaligned incentives if it doesn't.
There's something faintly unsettling about the prospect of AI agents conducting hiring conversations with other AI agents, each optimizing for different objectives. Perhaps that's the future anyway.
Margin Compression and Regulatory Shadows
If Perfectly's claims hold—same or better outcomes at half the cost—the margin compression for traditional recruiting agencies could prove severe. The industry has always operated on relatively fat fees justified by relationship capital and specialized knowledge. Strip that away with automation, and suddenly the economics shift dramatically.
But there are complications beyond competitive dynamics. New York City's Local Law 144, which took effect in July 2023 and remains in force, requires annual bias audits for automated employment decision tools used within city limits. The European Union's AI Act classifies recruitment AI as "high-risk," triggering conformity assessments and transparency requirements. Whether Perfectly's agents fall under these regulatory frameworks depends on implementation details the company hasn't made public.
Then there's the employer brand question. Companies obsess over how they present themselves to candidates—it's why they pay recruiters in the first place. Will hiring managers trust an autonomous agent to represent their culture, handle sensitive conversations, manage candidate expectations? The answer likely depends on the role, the company, and how much control Paul actually cedes to its human overseers.
A Crowded, Shape-Shifting Market
Perfectly enters a category experiencing both gold rush and consolidation simultaneously. Workday acquired conversational AI platform Paradox in August 2025, a signal that enterprise buyers see recruitment automation as strategic rather than experimental. A different company—Perfect, based in Israel—raised a $23 million seed round in early 2025 for recruiter co-pilot tools. LinkedIn, Gem, Findem, and applicant tracking systems like Ashby and Lever all have AI features in various stages of deployment.
The Winter 2026 YC batch, which presented at Demo Day on March 24, included nearly 190 companies according to post-event coverage from TechCrunch. Perfectly wasn't among the highlighted startups in that recap, though the batch itself reflects Y Combinator's continued appetite for AI infrastructure and vertical SaaS plays. Everyone, it seems, is suddenly building agents.
Table Stakes or True Edge?
Whether Paul can deliver on the efficiency claims remains an open question at this early stage. So does the deeper question of trust: will companies hand over a function as sensitive as talent acquisition to an AI that operates with meaningful autonomy? Perfectly's founders know machine learning. They've also experienced hundreds of agency-sourced interviews from the other side of the table. That experience might be a genuine edge—or it might just be table stakes in a category where competitive advantage now hinges on execution speed and data flywheel effects.
For now, Perfectly is in the market, taking placements, collecting performance data, refining the model. The founders are betting that recruiting, like customer support and sales development before it, will eventually succumb to automation. The agencies, presumably, are betting otherwise. Someone's going to be right.
