Maybe people need AI automation for tedious job searches?
I've been noticing a really common pain point lately talking to friends and seeing posts across different communities. So many people are spending 10+ hours every week just scrolling through job boards, reading through hundreds of postings, and filling out the same repetitive application forms over and over again.
It's incredibly draining, leads to major burnout, and a lot of people end up missing good opportunities simply because they can't keep up with the sheer volume of postings out there. They need a way to streamline this process without sacrificing the quality or personalization of their applications.
I've been thinking about a few different approaches to solve this. One idea is a simple Chrome extension that scrapes job boards as you browse, uses AI to evaluate how well each position matches your resume and preferences, and automatically fills out application forms for the ones that are a good fit.
Another angle is a centralized dashboard that pulls jobs from all the major boards in one place, scores them based on your specific criteria, and lets you apply to all your top matches with just a few clicks.
And then there's the more ambitious approach of a full AI career agent that learns exactly what you're looking for, proactively searches for new opportunities every day, and handles the entire application process from start to finish, including writing personalized cover letters tailored to each individual role.
I think there's a lot of potential here to make the job search process way less stressful and more efficient. Curious what you all think — does this resonate with you? Are there other angles to this problem that I'm missing?
Nice post
one thing worth flagging — theres a real risk of an arms race here. if job seekers all start auto-applying with AI, employers respond by adding more screening filters and AI detection, which makes the whole process worse for everyone. the tools that already exist for mass applying (like LazyApply etc) have already started triggering this on the employer side.
id actually argue the higher value play is helping people apply to fewer, better-matched roles rather than more. like Doron said, the matching layer is the real unlock. a tool that says "skip these 50 postings, these 3 are actually worth your time" saves more pain than one that blasts out 100 applications.
One angle worth exploring before jumping to auto-submission: let the user upload their resume, run it through an LLM to extract and generate the most relevant job titles based on their actual experience, then use those titles as the search terms against job boards filtered by postal code. Most people search too broadly or use the wrong titles for their skill set — the LLM can fix that mismatch before a single search even runs. The results immediately feel more meaningful because they're derived from what the person actually does, not what they think they should be searching.
For MVP, keep it pure search and matching — no submissions. Just show people "here are the roles that actually fit your background, near you." That alone solves a real problem and gets you signal on whether users trust the output before you build anything more complex.
The submission layer is where the real SaaS plays out. Gate it behind a subscription and build an approval flow where users review and greenlight each application before it goes out. That human checkpoint is actually the key differentiator — it keeps quality high, reduces the "spray and pray" problem, and removes the liability of fully autonomous submissions that could misrepresent someone. It's also a much easier sell: you're not replacing the job seeker, you're doing the legwork while they stay in control of what goes out with their name on it.
If you go the Chrome extension route, LinkedIn / Indeed / Greenhouse DOMs change weekly and selector-based scrapers break constantly — worth wrapping the extraction in a small LLM pass on the raw job posting HTML instead of relying on brittle CSS selectors. For the matching score, embedding the resume + each job and using cosine similarity outperforms keyword matching by a lot, but the real lift is letting users tune a few weights (seniority, remote, stack) on top of the embedding score rather than treating it as one black-box number.
this need more research into. Job emails and application can be really discouraging and tiresome