A Human Review

Mid-year reviews have just come around in the ol' 9-5. It’s an opportunity to reflect on the first six months (H1 if you’re into all that business speak) and set some goals for the end of the year. I personally find the process quite cathartic. Not only does it make the end of year review simpler (I can barely remember what I had for breakfast, let alone what business metrics I contributed towards in January), I find that as I start reflecting & writing, it quickly becomes a gratitude list of the smart colleagues I’ve worked with, and the interesting features I’ve been able to work on.

I spent a couple of hours working through my notes and writing up a review before sending it to my manager, awaiting their equivalent review for me.

However, the review I received back was, quite simply, quintessential AI slop. All the tell-tail signs were there—words like “pioneered” & “spearheaded” peppered throughout, no personal detail, and most of all, a complete regurgitation of what I originally wrote in my review. You see, managers ask to see your review before you submit so “we can submit a unified message to the business”, and you can quite quickly join the dots on what happened next (hint: it involved a bot called Claude & some cmd-c and cmd-v).

Without any context of me as a person, AI decided the best way to set personalised goals for me, was to fire back what I’d written but on steroids. When I wrote about “component migrations ahead of a React 19 upgrade”, AI wrote: “Trys should spearhead the React 19 upgrade project”, missing the fact that the project already has a feature lead. When I wrote about having an interest in improving our frontend stack, AI wrote “Trys should write a frontend strategy document for the company” - not exactly a task for a lone engineer, eh? There was no insightful feedback or contextual suggestions, just an A4 page of vagueries.

This is not a pop at the individual manager. I get it, if you have 15 reviews to deliver in a week and there’s a tool that takes all that pain away, that tool’s going to look pretty seductive. And I know it’s not just the managers. When the review deadline looms and you’ve not had a minute to reflect, let alone write, I’d wager a hefty sum that a good proportion of employees are submitting their reviews from bullet points via an OpenAI server.

But when you get to the point where both parties are using AI, you might as well pull the facade and send bullet points to one another. I’d rather read that than a wall of slop. I’d rather read the prompt. And I’d take a single ounce of humanity over a full time AI manager, any day of the week.