AI for employees
AI is coming for your task list. Not for you.
I’m not going to tell you nothing changes, because plenty does, and you’ve read enough LinkedIn posts pretending otherwise. I run my own operation with an AI crew, so I can tell you exactly which parts of a job it takes, which parts it can’t touch, and how to end up on the right side of that line. Whether you stay in your job or take it somewhere better.
What actually leaves your job
The tasks that go first are the ones with a defined input and a checkable output: first drafts, formatting, collating, summarising, chasing, reconciling, the weekly report nobody enjoys writing. In my crew those jobs run on a schedule and produce drafts for review. That’s not a prediction, it’s what already happens on my machine every night.
Be honest with yourself about how much of your week that covers. For most office roles it’s a real chunk, and pretending otherwise is how people get blindsided.
What stays stubbornly human
Judgement calls nobody scoped in advance. Relationships: the client who trusts you, the colleague you can read in a hallway. Accountability, because when the work is wrong, a person answers for it, and no vendor anywhere will take that on. Knowing what good looks like well enough to catch the plausible draft that’s subtly wrong.
Notice the shape of that list. It’s the senior half of every job. The grind leaves, the judgement stays, and the person who can do the judgement half well becomes more valuable, not less.
The move
Become the person who directs the machine
Every workplace adopting AI needs someone who can run it properly, and almost nobody can yet. That skill set is specific, learnable, and nothing like “knows how to prompt”:
Writing a sharp brief
An agent is only as good as its brief: what the job is, what it can touch, what done looks like. People who can spec work clearly were always valuable. Now it’s the difference between AI that helps and AI that flails.
Reviewing against a standard
AI output is confident whether it’s right or wrong. The person who can read a draft against the brief and catch what a machine missed is the review layer, and the review layer is where the responsibility lives.
Knowing where the limits are
Real autonomy is bounded: defined jobs, on a schedule, reviewed before they matter. Knowing what to hand over and what to keep is a judgement skill, and you build it by using the tools, not by reading about them.
Keeping the memory
The hard part of working with AI isn’t the chatting, it’s the context: the tidy record of how your team works that lets the tools start warm instead of cold. The person who owns that knowledge base owns the leverage.
Questions to ask your employer
I keep a list of questions to ask anyone selling AI. Here’s the employee’s version, and you’re entitled to straight answers:
What’s the plan for AI in my role?Not the company deck. Your role, your tasks. If nobody has thought about it, you’ve just learned you should think about it first.
Do the freed hours become better work or fewer people? An employer with an honest answer, either answer, is one you can plan around. An employer who dodges it has answered too.
Who reviews what the AI produces?If the answer is “nobody, it’s automatic”, the organisation doesn’t understand the tools yet, and the review job is sitting there unclaimed. That could be good news for you.
Will I get access and time to learn it? The companies worth staying at treat this like any other capability and train for it. The ones that expect you to compete with AI rather than run it are telling you where this is headed.
If you’re changing jobs, or thinking about it
The strongest position in the current market is being able to show, not claim, that you can direct AI on real work. A candidate who can say “here’s a job I automated, here’s the brief I wrote, here’s how I reviewed the output” walks past a stack of CVs that list “AI” as a skill. You don’t need your employer’s permission to build that evidence: the real tools run on your own machine, on your own account, on your own time.
And a caution from someone who sells in this market: most of what is marketed as “AI skills” is a costume, a PDF of prompts with a lanyard on. Before you spend money or a weekend on anything, including mine, read the truth about hiring an AI employee. It’s written for buyers, but the reality test in it works just as well on courses and gurus.
The honest bit
What I won’t promise you
AI won’t do your job for you, and nothing here says it will. Learning the real tools takes real effort: this generation of them is developer-shaped, and week one involves a terminal, config files and feeling briefly stupid. Some roles will shrink no matter how well the people in them adapt, and anyone who promises otherwise is selling something. What I can tell you is what I see from inside my own operation: the leverage went to the person directing the crew, and that seat is still mostly empty in most workplaces. It’s claimable. That’s the honest pitch, and the whole of it.
Start with the truth, not the tools.
The free guide is the reality test for everything sold with “AI” on the label, jobs advice included. Read it before you spend anything on anyone. Me included.