I don’t know who needs to hear this, but if you’re using AI for work, chances are you’re already a manager, whether you wanted the job or not.
Working with AI agents has made management part of the job even for people who have never had direct reports. Designers, engineers, researchers, and any professional who has managed to add (sometimes shoehorn, because they were asked to) AI into their workflow, are now delegating work, setting expectations, defining guardrails, and reviewing results they didn’t produce themselves.
Doing those right requires management skills, and alongside professional expertise, they determine how much value you get from AI. An agent producing a draft in thirty seconds means very little if reviewing and correcting it takes another two hours.
Don’t get me wrong, AI has been invaluable in helping me optimise my workflow. It allows me to iterate and deliver value faster than ever. I am absolutely not against its usage, on the contrary – prompt away, but don’t expect it to do your work for you or be anxious that it will replace you as a professional. Your agent only has as much expertise as the amount of context you provide it with.
That is why, if you feel like your role is soon to be replaced, learn good management skills, along with prompting.
Execution has to be preceded by good delegation. Agents need clear objectives, relevant context, guardrails and a standard to work toward. You need to decide what they can handle independently, what requires review, and where it should absolutely get you involved. Those boundaries should reflect the consequences of getting something wrong or misdirected.
“Use your judgment” is a vague instruction when you haven’t established what matters.
Your expertise gives AI instructions substance. It helps you identify the difficult or the important parts, anticipate mistakes, and evaluate the result. A convincing answer can still rest on a weak assumption. Recognising and knowing when to challenge that requires proficiency in the job (and that is where you are still irreplaceable).
Research is a useful example. An agent can gather information, compare sources, and produce an analysis. But you may hear hesitation in a customer’s answer, recognise a workaround they’ve stopped questioning, or understand why a stakeholder’s agreement should be treated with caution.
Some of those nuances you catch, might come from context the agent has no access to, or purely from your professional intuition. It is on you to bring those into the work, the agent never will. You know which questions deserve attention and which apparently sensible recommendations will fall apart in practice.
“Nah, I have years of management experience, I got this.”
Awesome! You most likely do. Are you sure though? Because AI can absolutely make you a worse manager. If the agent defines the problem, chooses the approach, and sets the criteria for success, you can end up approving work entirely on its terms. Each step looks reasonable, but your expertise has stopped directing the process. You want an agent to challenge your thinking, but you still need to understand why you’re accepting its advice.
I’ve seen this repeatedly with e-commerce store owners I’ve worked with. They turn to AI for direction on their business, design and brand decisions, then come to me saying, “Something isn’t right, but we’re not sure what.” I’ve also worked on products where I’ve had to argue my way through AI-generated requirements that weren’t human-led or even proofread. At that point, part of your job becomes getting people to examine requirements they’ve already accepted. That takes time they assumed had been saved by removing the human from the loop.
For senior leaders, that responsibility carries further. An AI suggestion can become a team’s priority as soon as the CEO endorses it. Before committing people’s time, the recommendation needs scrutiny from those who understand the customers, the systems, and the constraints. Their experience deserves weight. The more authority you have, the more carefully you should examine advice before turning it into someone else’s work.
I expect the strongest professionals to become very good at this combination: knowing their craft, delegating deliberately, and staying involved in the decisions that require their judgment. Clear instructions, sensible guardrails, and rigorous review deserve practice alongside technical skills.
We will be directing more work that we don’t personally execute. That makes our responsibility for its quality increasingly important.
If my name is on the work, I need to be able to stand behind it.
