'I vibe-coded a career review platform and didn’t cry once'

There’s a persistent stereotype that HR people don’t do tech. That we love a spreadsheet, sure, but we draw the line at anything involving code. Or automation. As for AI, we don’t know her.

I’ll admit, I’ve sometimes leaned into this stereotype. After all, I’m a middle-aged woman with two cats and a humanities degree. I’d proudly consider myself a fluffy, ‘people-person’ which is why I look after HR and research, not product or engineering. I still don’t fully understand what an API is and I’m now too embarrassed to ask.

However, the reality is I work at Nous. And at Nous, that excuse doesn’t really fly.

AI isn’t someone else’s job

We’re building one of the UK’s most ambitious consumer AI businesses. Our product is an intelligent agent - one that actively manages your household bills on your behalf - and our whole company is being built in its image.

AI at Nous isn’t a bolt-on; it’s the engine itself. Every team here uses  this new class of tools to work smarter, move faster and build better systems. From customer ops to facilities and finance, we all use AI to augment our work. When we say we’re an AI-native company, we don’t just mean the product. We mean the people too.

Which is how I (someone who experiences palpable fear every time I have to create an account on a new platform), ended up vibe-coding a fully automated career development workflow using Google Apps Script, Gmail and ChatGPT.

To my astonishment - and everyone else’s - it worked.

From CharlieHR to Lyd.ai

For a long time, we ran performance reviews at Nous using a patched-together process: CharlieHR to collect inputs, Slack to nudge managers, Google Sheets to track responses etc. It worked, but it was clunky, costly and required a lot of manual admin. I found myself copying and pasting the same instructions into dozens of docs, or trawling inboxes to see who’d filled out what. It felt like a process stuck in the wrong decade.

So I decided to fix it. Back in May, our CEO Greg set everyone at Nous a challenge: deliver a project that wouldn’t have been possible without AI. I didn’t know it then, but I already had the ideal candidate. We were barrelling towards our next round of 360 performance reviews and I wasn’t relishing the prospect of administering the process, so the timing was perfect. 

The initial goal was fairly modest: simply automate the boring bits. Create the folders, copy the templates, send the right instructions to the right people at the right time and (crucially) don’t break anything. But as I started sketching it out, I realised this wasn’t just about automation. It was a chance to redesign the whole workflow to make it structured, agentic and low-code, so that it could scale with us.

And so that’s what I built. Or more accurately, what I and my very able assistant ChatGPT built. Yes, this project consumed most of my evenings for the better part of a fortnight and I blew through all my o3 tokens and my lexicon of expletives pretty quickly. But my new system handles every stage of our review cycle - from creating Docs to logging statuses - with barely any manual input. It runs entirely on our own rails (Docs, Sheets, Gmail, Apps Script), and saves me hours of admin - not to mention thousands of pounds on some off-the-shelf SaaS platform.

More than that, it’s become a quiet proof-of-concept. A demonstration that the AI learning curve is climbable, even for Gen X laggards like me. It’s a sign that Nous’s AI-native culture isn’t just rhetoric; it’s real and it’s working, even in the corners of the company where you might least expect it. 

Lessons from a reluctant technologist

I didn’t set out to learn anything profound - I just wanted fewer tabs open. But along the way, I did pick up a few lessons:

• You don’t need to speak fluent Python to build something useful. But you do need to understand the workflow better than anyone else.
• ChatGPT is a brilliant colleague. Until it starts hallucinating about modal windows that just DO NOT EXIST. Then it’s like dealing with an overtired toddler (which, to be fair, I’m quite experienced at).
• There’s always a cleaner way. If something feels faffy, it probably is.
• Systems thinking is a slippery slope. One minute you're automating folder creation, the next you're explaining logic trees at dinner parties. I now dream in if/else statements and JSON objects. Frankly, I’m a stranger to myself.

Most of all, I’ve learned that the line between ‘technical’ and ‘non-technical’ is blurrier than it looks, and it’s getting blurrier by the day.  As an HR person, I cannot tell you whether AI will eventually take your job, but – as the Linkedin truism goes – someone who knows how to use AI almost certainly will. Like it or not, it’s time to get with the programme.

Why this matters (beyond HR)

At Nous, we don’t do middle management. We don’t hire project managers to talk about the work. We hire people who do the work and we trust them to figure things out, even when it’s messy, new or involves JavaScript.

That’s part of the culture we’re building. AI-native not just in our tools, but in our instincts. A team of people who can sketch a better system, spin up a prototype and ship it. Even (or perhaps especially) if it’s outside their official job title.

So if you’re reading this and wondering whether AI is for you, here’s my honest take: if a cat-owning, poetry-degree-holding HR lead can vibe-code a review platform from a drag start in about a fortnight (around her day job), you’ll probably be just fine.

And if you do end up chatting to me in an interview (we have lots of great roles open - take a look), just know this: I still don’t understand what an API is. Please don’t ask me.

‍

‍