Paul the Recruiter Is an AI, and He Wants to Fill Your Open Roles Before You Finish Your Coffee

The Macro: Hiring Is Slow, Expensive, and Everyone Knows It

Recruiting has a dirty secret: the process hasn’t meaningfully changed since job boards replaced newspaper listings. You post. You wait. A recruiter (human, expensive, commission-hungry) sends you a stack of resumes. Half of them are wrong. You waste two weeks. Repeat.

AI adoption in HR tasks jumped to 43% in 2025, up from 26% the year before, according to SHRM. That’s a big number, but it mostly means companies are using AI to write job descriptions and schedule interviews. The actual sourcing-to-hire pipeline is still painfully manual in most places.

The staffing market context is interesting here too. The US staffing industry is forecast to grow a cumulative 10% between 2025 and 2030, per Staffing Industry Analysts, even as revenue dips slightly before rebounding. That’s a market that’s growing, but also one that’s showing cracks. Traditional agencies are expensive and slow. ATS tools like Greenhouse or Lever are infrastructure, not intelligence. They track candidates; they don’t find them.

The crop of AI recruiting startups trying to wedge into this gap includes things like AI sourcing overlays, automated outreach tools, and candidate screening bots. None of them have fully collapsed the funnel into something that feels like a single product. They’re modules, not agents.

The interesting bet right now is whether a startup can build something that actually replaces the workflow end-to-end, not just one step of it. Your resume being a lossy format is part of the same problem: every layer of the hiring process introduces noise. Whoever can reduce the noise across the full stack probably wins.

That’s the space Perfectly is trying to occupy.

The Micro: One Agent, Your Slack, Interview-Ready Candidates

Perfectly describes itself as an AI-native recruiting agency, and the framing is deliberate. It’s not a platform you log into. It’s not another ATS. The product is closer to outsourcing your recruiter to a software system that then reports back to you inside Slack.

The agent has a name: Paul. Paul does sourcing, outreach, screening, and qualification. You get candidates dropped into Slack when they’re ready for an interview. The pitch is that this compresses a process that normally takes weeks into days.

The technical angle is worth paying attention to. According to Pendium AI, Perfectly was built by ML scientists from TikTok and Meta. The LinkedIn profiles confirm this more specifically: Zhuang (Gary) Luo is CTO and an ex-MLE at what appears to be TikTok, and Huimin Xie carries the title CAIO (Chief AI Officer). Victor Luo is CEO. All three are listed as co-founders and the company is part of YC’s W26 batch. Victor posted publicly on LinkedIn about how he found his co-founders for YC W26, which is a small but telling signal: this is a team that formed with intention around a specific problem.

The TikTok recommendation ML background is not a throwaway credential. TikTok’s recommendation systems are genuinely some of the best in the industry at matching content to user interest at scale. Applying that pattern-matching instinct to candidate-to-role fit is a real insight, not a resume flourish.

The “white-glove treatment” language for candidates is interesting too. Most AI recruiting tools optimize entirely for the employer. Perfectly is at least claiming to care about candidate experience, which matters for close rates once you’ve found someone good.

It got solid traction when it launched, which tracks for a YC company in a space this hot.

The thing I’d want to poke at: what does Paul actually do when a role is niche or weird? Sourcing for a senior Rust engineer at a crypto startup is a different problem than sourcing for a growth marketer. Whether the agent degrades gracefully in edge cases is unknown from the outside.

The Verdict

I think this is a legitimately interesting bet, with one big caveat.

The core product decision, putting an AI agent inside Slack and making it feel like a human recruiter rather than a dashboard, is smart. It meets the customer where they already live. Startups don’t want more tools. They want fewer.

The TikTok ML background is real differentiation if it actually shows up in match quality. That’s the number that matters. Not how fast Paul sends candidates, but how often those candidates convert to hires. A faster firehose of mediocre candidates is worse than the status quo.

At 30 days, I’d want to know their fill rate on the first few customers and whether those hires are surviving past 90 days. At 60 days, I’d want to know if Paul handles edge-case roles without human fallback. At 90 days, I’d want to see whether they’re retaining customers past a single hire or whether companies are treating it as a one-time tool.

The YC stamp and the team background are real. The risk is that recruiting quality is slow to measure, and “interview-ready candidates in days” is a claim that takes months to validate in the market. If the first cohort of customers get good hires, Perfectly has something. If Paul turns out to be fast but imprecise, the agency framing becomes a liability fast.