<aside> 👋🏻
I’ve led People teams across countries and today I specialise in People Automation, using AI, automation, and practical experiments that turn ideas into impact.
This article is part of my No New Tool Required series, where I share smart, lightweight experiments to free up People team time and generate better insights with the tools we already have.
🔗 Let’s connect on LinkedIn
</aside>
I'm tired of the word "agents."
Every workflow is suddenly an agent. Every automation gets rebranded as agentic. Every vendor pitch implies that if you're not deploying autonomous systems, you're already behind.
And it's making People teams second-guess automations that work perfectly fine — and overengineer solutions that don't need overengineering.
The AI space has collapsed three very different concepts into one buzzword.
Automation, agentic workflows, and agents are not the same thing. They represent different levels of delegation, different amounts of judgment, and different kinds of risk.
Treat them as interchangeable and you get:
The question isn't "Can we build an agent?" We can. Anyone can.
The real question is: What level of delegation does this work actually require?
Because most HR processes don't need maximum autonomy. They need the right level of autonomy matched to the complexity of the task.
Sometimes that's an agent. Often it isn't.
In this article, I'll break down the delegation spectrum I use when looking at a People Ops process. We'll use onboarding as the example — because we've all lived through it, and it shows why one workflow might require three completely different approaches.
Light on theory, no vendor pitches, focused on the simplest thing that could actually work.