AI did not make me better at my job. It made me faster at the parts that were never the hard part.
There is a version of this article that opens with something like “AI is transforming the workplace as we know it.” I am not going to write that one.
I want to write the version that is actually useful, from someone who works in administration and research support at a university, leads BriefDesk Solutions, and has spent the better part of the last year integrating AI tools into nearly every layer of how I work.
Here is what I have learned.
Most people adopt AI tools the wrong way
The typical adoption story goes like this: you see a demo, you are impressed, you start using the tool for everything, and three weeks later you have half-finished outputs you do not fully trust and a vague sense that something has gone wrong.
The problem is not the tools. The problem is that we treat AI adoption like software installation rather than skill development.
I have used AI tools to draft financial trackers, generate prediction models, and structure complex reports. In every case, the tool got me to a starting point. What came after was entirely mine. Each output required me to know enough about the domain to catch what was wrong, redirect what was weak, and own what came out the other end.
That ownership part matters. If you cannot explain the document you submitted, you should not have submitted it.
The productivity gains are real, but they live somewhere specific
People talk about AI productivity like it is a flat multiplier across everything you do. It is not.
Where AI tools genuinely help me: first drafts of structured documents, reformatting data, building frameworks I then reshape, summarising long inputs, generating options when I am stuck. These are tasks where the hard part is starting or organizing, not deciding.
Where AI tools do not help as much: anything requiring institutional memory, political judgment, relationship context, or domain nuance that has not been explained to the tool. My work involves committee governance, quality assurance processes, and cross-institutional coordination. The tool does not know the history. I do.
This is not a limitation to complain about. It is a boundary to understand.
The real skill is knowing when to stop the machine
One of the more uncomfortable things I have noticed: AI tools are very good at producing confident-sounding text that is structurally complete but contextually thin. They will write you a report that reads well and misses the point.
The skill you actually need is not “how to prompt.” It is the ability to read output critically enough to know when it has gone off track, and why.
That requires deep familiarity with your own work. You cannot delegate quality control to a tool you are using to improve quality control. Someone still has to hold the standard.
In my role, that person is me. In your role, it is you.
What this means for how we develop people
I work in a learning and teaching environment. We think a lot about how people develop capability, not just how they acquire tools.
AI literacy is not the same as AI fluency. You can know how to use a tool without understanding what it is actually doing. In professional contexts, that gap shows up eventually, usually at the worst time.
What I would advocate for is building AI use into professional development in the same way we build data literacy or communication skills: deliberately, with reflection, and with honest attention to where judgment is still required.
The question is not whether your team should be using AI tools. They probably already are. The question is whether they are using them with enough understanding to catch the gaps.
My actual position
AI tools have made me faster at a specific class of tasks. They have not made my judgment better. They have not replaced the need to understand what I am producing or why it matters.
I do not think they will. The value of experienced professionals is not that we produce documents quickly. It is that we know which documents to produce, what to include, what to leave out, and what the implications are.
That has not changed. And for now, I do not think it is about to.
First published on LinkedIn on 19 May 2026. Read the original.