JobPiloteer

Job searching is a brutal time sink: 30–60 minutes per tailored application, weeks of waiting, mostly silence. The application-to-interview conversion rate in most industries hovers around 2-5%, meaning 95%+ of effort is wasted. We set out to compress that time while improving those odds.

The product turned a 30-60 minute tailored application into a 2-3 minute task — without sacrificing the personalization that actually drives interview rates.

For users: they could submit to 5-10× more positions in the same time they used to spend on one. Each application was still customized to the job description and the applicant's CV; the agent handled the personalization work in the background that humans would otherwise grind out manually.

The system tracked outcomes (interviews, offers, rejections) and learned from response patterns to improve recommendations for future applications. Built on AI agents that could reason across multi-step workflows — not just chat-style single-prompt completions — at a time (2022) when this kind of orchestration was still uncommon.

Within 90 days of alpha launch, the aggregate interview rate across alpha users reached approximately 10% — roughly 10 interviews per 100 applications submitted through the platform, compared to a 2-5% industry baseline for cold applications. The product reached $10K MRR within the same window with 95% satisfaction scores in user surveys.

The compression of effort was the more striking metric: applications that previously took 30-60 minutes of manual tailoring shipped in 2-3 minutes, multiplying each applicant's reach by 5-10× in the same time budget.

**What this work demonstrates:** The agent architecture we built — multi-step reasoning, outcome tracking, learning from response patterns — is the foundation of every modern AI agent system. Building this in 2022 (when "AI agents" wasn't yet a buzzword) required designing primitives that became standard a year later.