How CVPilot Builds Trust Through Transparency to Grow User Adoption

A close mate of mine has been an engineer for nearly 8 years. Solid CV on paper. He applied to about 40 jobs over 3 months and got 2 interviews. We sat down one evening to figure out what was going on. His CV wasn't bad; the issue was that every job ad wanted slightly different things and he was sending the same CV to everyone. We spent maybe 90 minutes manually rewriting it for one specific role, and he had a callback that week. That was the bit that stuck with me. The skill wasn't the problem; the bridge between his experience and the application was. Every job seeker is rebuilding that bridge from scratch on their own. I started CVPilot the next weekend. The first version was a tiny script that diffed a CV against a job description and told you what was missing. Looked terrible, no UI, just a text output, but the logic was there. Built the proper version over the next couple of months in evenings and weekends.

Worked, scheduling proper blog content across LinkedIn, Dev.to, X, Facebook, and Instagram with Postiz. Not generic SEO fluff, actual posts about how ATS parsers work, why CVs get rejected, what recruiters actually scan for. One post about why 75% of UK CVs get filtered drove 13 LinkedIn organic sessions in a week from basically zero. Slow build but it compounds, and I now have around 90 posts scheduled across the 5 platforms. Flopped, cold-DMing job seekers on LinkedIn offering them free Pro access. Got near-zero replies and a couple of polite blocks. Honestly, it felt slimy when I was doing it, and apparently it read that way too. People don't want to be sold to in their inbox by a stranger, especially about something as personal as their career. I stopped that within a week and switched everything to publishing useful stuff where job seekers already are and letting them come to me. Slower curve, but the people who do come in are way more engaged than the ones I was harassing in DMs.

The biggest one early on was that people didn't trust the rewrite. They'd run their CV through, look at the output, and not actually use it because they couldn't explain in an interview why a particular bullet was changed. A few users straight-up told me they felt nervous shipping something they didn't fully understand. So I rebuilt the entire output around an audit trail. Every rewritten bullet now shows the original, the new version, the keyword that was added, and a one-line explanation of why we changed it. They can revert any single change or edit it inline in the resume builder. That fixed it. Support tickets dropped, people started actually exporting and using the rewrites. The lesson stuck with me, AI output people can't explain back to themselves is worse than no AI output at all. You have to show your working. Same principle now drives how we build everything else, the offer comparison shows reasoning behind each recommendation, the interview prep links each suggested answer back to a specific bit of the candidate's own experience.

Most people building in this space think CV tools compete on output quality. They don't. They compete on trust. Job seekers are anxious, often skint, applying to jobs that genuinely matter for their livelihood. They will not paste their CV into a tool that feels generic or produces stuff they can't defend in an interview. The actual moat isn't smarter prompts or fancier templates; it's whether the user feels in control of every word ending up on the page they're submitting. Tools that automate too aggressively get used once and abandoned. Tools that show their working, let users edit freely, and explain every change get used 20 times in a single job hunt. Everyone is racing to take humans out of the loop. The actual win is the opposite, making the human's input 10x more efficient. That insight is basically why the product looks the way it does, and it's also why I'm not worried about big AI players coming into this space. Speed of model improvement isn't what wins here; the relationship with the user does.

Push CVPilot to 2,000 paying Pro and Power subscribers within 12 months. To get there, I need to land 10 academy and bootcamp partnerships across the UK and Europe, where graduates are at exactly the moment we solve for. Already in conversations with one. Also opening up the offer comparison and interview prep as standalone entry points alongside the free ATS checker, and looking at a recruiter-facing API so hiring teams can run candidate fit audits at scale. Bigger picture, I want CVPilot to be the layer between any candidate and any application, regardless of where they found the job.

8 years writing production software end to end. Worked on early-stage SaaS startups where I was one of four engineers, and on enterprise apps that had millions of users hitting them daily. Mostly full-stack, comfortable across Next.js, Node, Python, Postgres, AWS, Stripe, BullMQ, AI integration. CS degree and a marketing MSc, which is a weird combination but actually maps surprisingly well to running a SaaS solo. CVPilot is the first product I’ve founded myself, though, so the engineering side has been the easy part. The rest of running a business - distribution, sales, partnerships, growth, content - I’m learning as I go, mostly by getting things wrong first and then figuring out what should have happened. The marketing degree helps with the framing side but doesn’t replace the daily grind of actually doing it.

Spent the first eight weeks over-engineering the AI pipeline before talking to a single real user. Built dual-model fallback logic, prompt caching, a self-critique loop, all before validating that anyone actually wanted what I was rewriting their CVs into. When I finally shipped and watched users react, the real friction wasn't AI quality; it was trust and editorial control. People didn't trust the output and wanted to see the reasoning behind every change. I'd solved the wrong problem first. Lesson, ship the thinnest version that produces real user reactions, then let those reactions tell you where to invest engineering time.