You’ve tried AI. You typed “write me a performance review” into ChatGPT. You got a generic, lifeless paragraph that sounded like it was written by an HR chatbot from 2014. You closed the tab and thought: AI is overrated.
Here’s the thing — you’re right. That output was terrible.
But the problem wasn’t the AI. The problem was you.
You just gave an assignment with no context about the person, no examples of what good looks like, no information about your review format, and no direction on tone or purpose. If you gave that same briefing to a new hire on their first day, you’d get the same garbage. And you wouldn’t blame the new hire — you’d blame yourself for a terrible handoff.
AI doesn’t have a performance problem. You have a management problem.
And that’s actually good news — because you already know how to fix it.
AI is a junior employee. Manage it like one.
Stop thinking of AI as a tool. Tools do what you tell them. A hammer hits a nail. A spreadsheet calculates a formula. You don’t need to “manage” a tool.
AI is different. It interprets. It makes choices. It fills gaps with assumptions. It produces output that varies depending on how you communicate with it. It can be brilliant or useless depending entirely on how you direct it.
That’s not a tool. That’s a team member. Specifically, it’s the fastest, most eager, most inexperienced team member you’ve ever had:
- Fast and tireless. It produces first drafts in seconds, never gets tired, never complains.
- Confident but often wrong. It will state incorrect things with absolute certainty — just like a junior who doesn’t know what they don’t know.
- Context-dependent. Give it great context, get great output. Give it nothing, get nothing useful.
- Zero institutional knowledge. It doesn’t know your team, your company culture, your boss’s preferences, or why that particular client is sensitive.
- Incapable of judgment. It cannot read the room, navigate politics, or decide when to break the rules.
Sound like any new hire you’ve ever managed? You already have the frameworks for this. Let’s apply them.
Framework 1: Brief it like you’d brief a new hire
The single biggest reason managers get bad AI output is bad briefing. It’s the same reason new hires produce bad work — not because they’re incapable, but because nobody told them what “good” looks like.
Bad delegation to a person: “Handle the client report.”
Bad delegation to AI: “Write me an email to my team.”
Both fail for the same reason: zero context.
Good delegation to a person: “The Q2 client report is due Friday. Here’s last quarter’s format. Focus on engagement metrics because Sarah’s team is watching those numbers. Flag anything that looks off — I’d rather catch it now than in the meeting.”
Good delegation to AI: “I need to email my team of 8 about a schedule change starting May 1st. We’re moving from 5-day to 4-day weeks, but some people will lose their preferred shift. Tone: honest and direct, not corporate. Acknowledge the inconvenience. Explain that this is driven by budget constraints, not performance. Keep it under 200 words.”
The difference? Five things:
- Who — who is this for (team of 8, specific client, your boss)
- What — what specifically do you need (email, review, talking points, analysis)
- Why — what’s the context and purpose
- Tone — how should it sound (direct, empathetic, formal, casual)
- Constraints — length, format, what to include or avoid
This isn’t “prompt engineering.” This is delegation. The same skill you use with every task you hand off to your team.
Framework 2: Set authority levels
One of the most useful management frameworks for delegation is the Authority Levels model:
| Level | Person | AI |
|---|---|---|
| 1 — Do exactly this | Follow these steps precisely | ”Write exactly this email, using these bullet points” |
| 2 — Research and recommend | Give me options with your recommendation | ”Give me 3 approaches to this conversation and tell me which you’d recommend” |
| 3 — Recommend, then act if I agree | Proceed unless I say otherwise | Not appropriate for AI |
| 4 — Act and report | Do it and tell me what you did | Not appropriate for AI |
| 5 — Full ownership | Own it completely | Never appropriate for AI |
With a junior employee, you start at Level 1-2 and work up as trust builds.
With AI? You stay at Level 1-2. Permanently.
AI should draft, research, recommend, and prepare. You should decide, edit, send, and own. The moment you let AI operate at Level 4 or 5 — sending emails, publishing content, or making decisions without your review — you’ve handed your judgment to something that doesn’t have any.
And unlike a junior employee who might come to you when they’re unsure, AI will never say “I’m not confident about this.” It will just do the wrong thing confidently.
The rule: AI prepares. You decide. Always.
Framework 3: Review every output
When a junior hands you their first draft of anything — a report, a presentation, an email — what do you do?
You review it. You don’t forward it to your boss without reading it. You check the facts. You adjust the tone. You add the nuance that only comes from knowing the people and the situation.
But I watch managers copy-paste AI output directly into emails, reports, and Slack messages every day. No review. No editing. No judgment applied.
That’s not a productivity hack. That’s negligence. You just outsourced your judgment to something that has none.
Your review checklist for any AI output:
- Is it factually correct? AI hallucinates. Verify any specific claims, data points, or references.
- Does it sound like me? If your team would read it and think “this doesn’t sound like my manager,” rewrite it.
- Does it account for context AI doesn’t have? The interpersonal dynamics, the political sensitivities, the history.
- Would I put my name on this? Because you are putting your name on it. If you wouldn’t sign a letter your assistant drafted without reading it, don’t send an email AI drafted without reading it.
Framework 4: Iterate — don’t quit after draft one
Here’s where most managers give up. They get a mediocre first response from AI and close the tab. “Tried it, wasn’t helpful.”
That’s like giving a junior employee one task, disliking the result, and firing them on the spot.
With a real junior, you’d say: “Good start, but the tone is too formal. Also, you missed the part about the timeline. Can you revise?” And they’d come back with something better. Maybe you’d go again. By the third round, it’s genuinely useful.
AI works the same way:
Round 1: “Help me prepare for a tough conversation with an employee who’s been missing deadlines.” → AI gives you a generic framework.
Round 2: “Make it more direct. This person has missed 3 deadlines in 2 weeks and I’ve already talked to them informally. This is the formal conversation. I want to be firm but not threatening.” → Much better. Specific, toned correctly.
Round 3: “Add an opening line that acknowledges I value their work overall — because I do. And end with a clear next step and timeline.” → Now you have something you can actually use.
Three rounds, ninety seconds total. The managers who say “AI changed how I work” are the ones who iterate. The managers who say “AI is useless” are the ones who quit after draft one.
Framework 5: Know its strengths and never delegate its weaknesses
Every good manager knows their team members’ strengths and blind spots. You give analytical work to the analytical person. You don’t put the introvert in charge of the client presentation. You play to strengths and compensate for weaknesses.
AI’s strengths — delegate these:
- First drafts. Emails, reviews, announcements, agendas, talking points. AI gets you 70% there in seconds.
- Preparation. Thinking through difficult conversations before they happen. Anticipating objections. Planning scripts.
- Perspective. “What am I not considering?” “Argue the other side.” “What would my team member think hearing this?”
- Structure. Turning your scattered thoughts into an organized plan, checklist, or framework.
- Research. Summarizing approaches, comparing options, providing starting points for your own thinking.
AI’s weaknesses — never delegate these:
- Judgment calls. Who to promote, who to put on a PIP, when to escalate to your boss. These require context AI doesn’t have.
- Relationship decisions. How to approach Marcus (who’s going through a divorce) differently than Sarah (who shuts down when surprised). AI doesn’t know your people.
- Reading the room. When to push and when to back off. When someone says “I’m fine” but clearly isn’t. AI can’t read body language or tone.
- Sensitive communications. Terminations, personal feedback, anything that lands differently based on delivery. AI can help you prepare, but you deliver it human-to-human.
- Final decisions. AI recommends. You decide. Every time.
Your weekly AI routine
Here’s what a practical AI-assisted management week looks like — no hype, no “10x productivity” nonsense:
Monday morning — Week planning:
“Here’s what’s on my plate this week: [list 3-5 priorities]. What should I be thinking about? What might go wrong? What could I delegate?”
You’re not asking AI to plan your week. You’re using it as a thinking partner to pressure-test your own plan.
Before any difficult conversation:
“I need to talk to [role, not name] about [situation]. Here’s what’s happened: [2-3 sentences]. I want to be [direct/empathetic/firm]. Help me plan: opening line, 3 key points, and how to end constructively. Then play devil’s advocate — how might they react?”
This is the single highest-value use of AI for any manager. Walking into a tough conversation prepared instead of winging it changes the outcome every time.
For any writing task over 200 words: AI writes draft one. You rewrite in your voice. This cuts writing time by 50-60% — not because AI writes well, but because editing is faster than staring at a blank page.
When you’re stuck on a decision:
“I’m deciding between [Option A] and [Option B]. Here’s what I know: [context]. Argue against each option. What’s the strongest case for the one I’m leaning away from?”
The counterarguments are often better than what you’d generate alone, because AI doesn’t have the same emotional attachment to your preferred option.
What you never do:
- Never let AI send anything without your review
- Never use AI output in a conversation without verifying facts
- Never put employee names, personal data, or confidential company information into AI tools
- Never ask AI to handle anything that requires empathy, relationship judgment, or knowledge of your specific people
The uncomfortable truth
Here’s what nobody in the AI hype cycle wants to say: AI doesn’t replace management skills. It amplifies them.
If you’re a good delegator, AI makes you faster. If you’re a bad delegator, AI makes you faster at producing garbage.
If you give clear context, set expectations, and review output — with people or with AI — you get great results. If you throw vague instructions over the wall and hope for the best, you get mediocrity. Every time.
AI is the ultimate mirror for your management style. If your AI results are consistently bad, the problem isn’t the technology. The problem is your inputs. And fixing your inputs — giving better context, setting clearer expectations, reviewing before sending — will make you better at managing AI and better at managing people.
That’s not a coincidence. It’s the same skill.
Start this week
Pick one thing from the weekly routine above. Just one.
If you have a difficult conversation coming up, use the prep prompt. If you’re procrastinating on a performance review, ask AI for a first draft with real context. If you’re stuck on a decision, lay out the options and ask AI to argue against each one.
Try it once. See if the output is useful. Edit it until it sounds like you. Then try it again next week.
Your newest team member is artificial. It’s fast, tireless, and eager. It’s also clueless, overconfident, and incapable of judgment.
You already know how to manage that. The frameworks are the same. The only thing that’s changed is the team member.
📖 New to AI as a manager? Start with the basics: How to Use AI as a New Manager — 7 copy-paste prompts for conversations, reviews, agendas, decisions, and more. Works in every industry, not just tech.
📚 Want to go beyond articles? See our Top 5 Books on AI for Managers — the books worth your weekend, chosen for people who lead teams, not build models. No hype.
🧠 How do you actually score? Take our free 4-minute quiz: Is AI Making You a Better Manager — Or a Lazier One? — 15 scenarios across Critical Review, Judgment Retention, Privacy & Safety, Skill Development, and Appropriate Use. Honest results, no email required.
The deeper read. The how to delegate as a new manager guide covers the same authority-levels framework applied to people, and the Delegation ROI calculator shows when it is mathematically worth handing something off — to a person or to AI.