Not everything needs AI. That's a completely valid position that nobody seems to be giving companies permission to hold.
And most don't even understand what AI is. AI isn't one thing. It's machine learning, natural language processing, large language models, amongst other things - different tools with different strengths and different limitations. Most organisations have been using some form of AI for years without thinking about it. Recommendation engines, spam filters, predictive text all leverage AI. It was embedded in the tools you already used, managed for you, and running quietly in the background.
What's changed is that it used to be controlled for you. Now you're the one controlling it. And these tools are more powerful, more autonomous, and non-deterministic; they're probabilistic, based on pattern matching and mostly a mirror of who we are. We treat the output as rational, when the source material often isn't.
GenAI and LLMs are trained on us. All of us - the brilliant research all the way to the shitposting Reddit threads. The peer-reviewed papers and the hot takes are all held in one space. AI doesn't know the difference, unless its trainers weigh this up.
There's also the question of accountability. When a person takes on a piece of work, they own it; they know they're responsible for it. AI doesn't have that. It's motivated to complete the task, in the most efficient manner possible. In high-stakes scenarios, AI working alone probably isn't your answer. Context matters enormously, and AI doesn't always know which context it's drawing from.
And while we're at it: automation and AI are not the same thing. If a task follows clear rules - the same input always produces the same output - just automate it mechanically. It's more reliable, more predictable, and cheaper. You don't need an LLM to tick a checkbox, though it feels more accessible.
AI is better suited to exploratory work: creative processing, navigating complexity, pattern recognition across messy information. Sometimes the most boring use cases are the best ones. Not everything needs to be a transformation story.
I use AI in my own work to work through complex problems at a depth and speed I couldn't match on my own. With proper governance, it elevates my ability to deliver - as long as I don't cognitively surrender. But for me, it's a thought sparring tool, not a delegation tool.
If your processes don't work, AI doesn't fix them. If your system isn't governed, AI just uses ungoverned data and knowledge. You're just automating the mess.
Get the foundations right first: the system, processes, the governance, the knowledge. Then AI has something solid to work with. Without that, you're adding an intelligent tool to a system that isn't ready for it.