There’s a small bot that applies to jobs for me now. It’s the least interesting part of this story.

It started the way every job hunt starts: too many openings, not enough hours. So I did the obvious thing first. I pasted a job description into an AI chatbot and asked it to tailor my resume to the role.

It worked too well.

To match the posting, it nudged my experience up a notch. Smoothed the edges. Made me sound a bit more senior, a bit more certain. Nothing it wrote was a lie, exactly. It was just a confident, fluent version of me that, even if I got shortlisted, I couldn’t back up in a room.

The chatbot wasn’t broken. It was being obedient. I’d asked it to get me past the filter, so it optimised for getting past the filter. It had no idea I’d be sitting across from a person a week later, owning every word. It solved the problem I typed. It never saw the problem underneath it.

I could have fixed that with a better prompt. Spell out the rules, tell it to stay grounded, and it behaves. But then I’d be running that prompt by hand for every single opening, reading every draft, catching every slip, forever. That’s the kind of discipline you run out of by Tuesday. And to be honest, I’m a little lazy. Building a machine to be diligent for me sounded a lot more appealing than becoming a diligent person. 😛

So I built the bot. It finds the openings, checks each one against my background, and only bothers tailoring a resume where I’m genuinely a fit; the rest it skips. The honesty lives in the pipeline now, not in my willpower, so a claim can’t drift far from something I’ve actually done. Then it sends me the ones worth applying to. (Getting it to reliably reach the jobs at all was its own fight; the modern web is walled off against bots, so I had to redesign it more than once. A story for another day.)

And it worked. The resumes came out true. They also came out sounding like everyone else’s.

Same fluent, weightless voice all these tools have. “Data-driven insights.” “Operational excellence.” “Cross-functional collaboration.” I’d fixed the lying and walked straight into the cliché. Which matters, because the recruiters reading these can spot that voice from across the room. They see a hundred a week. A truthful resume that reads “robot written” still loses.

That’s the strange part. My application gets written by a machine and then read by one, and somewhere in the handoff the actual person is supposed to survive. Mostly they don’t. The system loses at both ends. Recruiters, buried under near-identical applications, can’t find the right person in the time they have. And people who would genuinely fit get dropped over things as small as a missing keyword, a weak line of positioning, a KRA framed a little too broadly.

It’s a real problem. It’s also not one you fix by being angry at it. AI was supposed to make us sharper and take the boring weight off our hands, not bury us deeper in the data, the platforms, the noise. The tools most of us actually have, the public ones, not whatever the big companies run behind closed doors, have made this part harder, not easier.

So the skill worth having now isn’t a cleverer prompt. It’s adaptability: the willingness to keep changing your approach as the walls move, which, building this thing, I did more times than I’d care to admit. The bot didn’t make me a better candidate. It made me a faster learner.

I still use it. I also still read every line before anything goes out, because the one thing this taught me is that the moment you stop watching is the moment it starts solving the wrong problem, beautifully.