How to Use AI as a Manager (And When Not To)

Every article about AI for managers falls into one of two camps. The first says AI will transform everything, that you should be using it for every email, every meeting, every decision. The second says AI is overhyped and managers should stick with what works. Both are wrong.

The reality is more boring and more useful than either extreme. AI is very good at some management tasks and genuinely bad at others. The managers getting the most out of it are not the ones using it for everything. They are the ones who figured out how to use AI as a manager in the spots where it actually helps, and how to stay away from it where it does not.

This matters because the stakes are not theoretical. When you use AI to draft a performance review, the words it generates will land on a real person’s desk and shape how they feel about their work. When you use it to summarize meeting notes, you are trusting it to capture what actually matters, not just what was said. When you skip it entirely because you think it is a fad, you are spending two hours on something that could take twenty minutes.

The problem is that nobody teaches managers how to make this distinction. The tool companies want you to use AI for everything because that is how they grow. The skeptics want you to avoid it entirely because that is a safer opinion to have. Neither side has to manage a team of eight people while juggling performance reviews, weekly updates, and a project that just went sideways.

According to Gallup’s Q4 2025 workplace survey, 55% of managers have used AI at work, but only 30% use it frequently. Nearly half of all U.S. workers say they never use AI in their role. The adoption is happening, but the guidance on how to use it well has not kept up.

This article is the guide I wish I had when I started using AI in my own management work. It is not a list of tools or prompts. It is a framework for deciding when AI makes you a better manager and when it makes you a lazy one.

What AI Actually Does Well for Managers

The best use cases for AI in management share one trait: they involve turning raw information into a structured first draft. Not a final product. A starting point that would have taken you 30 to 60 minutes to create from scratch.

First Drafts of Recurring Documents

Performance reviews, weekly status updates, goal-setting frameworks, and meeting agendas. These all follow patterns. They have a structure, a tone, and a purpose that do not change dramatically from one instance to the next. AI is excellent at generating a first draft you can edit in five minutes instead of staring at a blank page for twenty.

The blank page problem is real. Most managers do not struggle with knowing what to say. They struggle with starting. AI eliminates that friction entirely. If you have never tried this, start with something like a ChatGPT prompt for performance reviews and see how much faster the first draft comes together.

The numbers back this up. Managers typically spend three to six hours per performance review gathering notes and writing feedback. Lattice’s 2025 State of People Strategy Report found that 49% of managers struggle to review a full year of feedback, and 42% find the review process a burden. Companies like Citi have responded by rolling out AI drafting tools. According to HR Dive, their Performance Assist tool generates first drafts by pulling data from internal systems, and less than one percent of employees have opted out of having their manager use it.

Summarization and Synthesis

If you have a long email thread, a dense report, or a set of meeting notes that need to be condensed into something digestible, AI handles this well. It can pull out key points, organize them logically, and present them in a format your audience can actually read. This is mechanical work. It requires no judgment about what matters, only the ability to compress information without losing meaning. That is exactly what large language models are built to do.

Rewriting for Tone and Audience

You wrote a blunt status update and need it softened for an executive audience. Or you wrote something too formal and want it to feel more conversational for your team. AI can shift tone without changing substance. This saves time and prevents the overthinking that happens when you try to rewrite something for the third time because it does not feel right.

Brainstorming and Option Generation

When you are stuck on how to approach a difficult conversation or need five different ways to frame a piece of feedback, AI is a useful thought partner. It will not give you the right answer, but it will give you options you had not considered. The value is in expanding your thinking, not in outsourcing it.

Research and Fact-Gathering

Need to understand a concept before a meeting? Want to compare how other companies handle a specific policy? AI can gather and organize background information quickly. It is not always accurate, which means you need to verify anything important, but as a starting point for research, it saves significant time.

The common thread across all of these is that AI handles the mechanical layer of management work. Formatting, structuring, drafting, compressing. These are tasks that require effort but not judgment. They take time, but do not benefit from your years of experience managing people. Every minute you save on a first draft is a minute you can spend on the work that actually requires you to be in the room.

The mistake happens when managers assume that because AI did a good job on the first draft, it can also handle the final version. It usually cannot. That is where the line starts to blur.

Where AI Falls Apart

AI fails in predictable ways, and most of them come down to the same root cause: it does not know your people.

Emotional Intelligence

A manager’s job is not just to communicate information. It is to communicate information in a way that lands correctly with a specific person in a specific moment. AI does not know that Sarah is going through a divorce and needs a lighter touch in her review. It does not know that Marcus responds better to direct feedback than encouragement. It does not know that your team is exhausted from a three-month sprint, and a cheerful “great work, team!” email will feel tone-deaf.

Context is everything in management, and AI has none of it unless you provide it. Even then, it processes context as data points. It does not feel the weight of them the way you do. This is especially true in difficult employee conversations where tone and timing matter more than the words themselves.

Institutional Knowledge and Culture

Every organization has unwritten rules. Who actually makes decisions. Which meetings matter and which are theater. What the real priorities are versus what the slide deck says. AI cannot learn these things because they are not written down anywhere. When you ask it to draft a message to your VP, it does not know that your VP hates bullet points and only reads the first two sentences. When you ask it to write a project update, it does not know that the real audience is not the people on the distribution list, but the one person who will forward it to the CEO.

Judgment Calls

Should you promote the stronger performer who is difficult to work with, or the slightly weaker one who elevates the entire team? Should you push back on an unrealistic deadline or find a way to make it work? Should you flag a struggling employee to your boss now or give them another month? These are not information problems. They are judgment problems. AI can lay out the pros and cons. It can even tell you what a management textbook would recommend. But it cannot weigh those factors against the specific relationships, politics, and history that make your situation unique. The answer always depends on things AI cannot see.

Authenticity

People know when something was written by a person and when it was generated. They may not be able to articulate how they know, but they feel it. A performance review that uses phrases like “consistently demonstrates” and “proactively identifies opportunities” reads like a template because it is one. Your team wants to hear from you, not from a language model wearing your name. This does not mean you cannot use AI to start the draft. It means the final version needs to sound like something you would actually say. If you would never use the word “synergy” in a conversation, it should not appear in your written communication either.

The Pattern

AI fails when the task requires knowing something that is not in the prompt: your people, your culture, your history, your instincts. The more human the task, the less useful AI becomes. Not useless. Just less useful. The problem is that the most essential parts of management are almost entirely human. Research supports this. Despite growing investment, 70 to 85% of AI initiatives fail to meet their expected outcomes, according to studies from MIT and the RAND Corporation. The technology works. The problem is applying it to tasks that require more than processing power.

The Decision Framework

Knowing that AI is good at mechanical tasks and bad at human ones is useful in theory. In practice, most management tasks are somewhere in the middle. You need a way to evaluate each situation quickly without overthinking it.

The Three-Question Test

Before using AI for any management task, ask yourself three questions. They take about ten seconds, and they will save you from the two most common mistakes: using AI when you should not, and avoiding it when you should.

Question One: Is This a First Draft or a Final Product?

If you are creating something that will be reviewed, edited, and shaped by you before anyone else sees it, AI is almost always a good starting point. Performance review drafts, meeting agenda outlines, project update templates. These benefit from AI because the output is raw material, not the finished work.

If the output will go directly to another person without meaningful editing, be careful. The email you send to a struggling employee after a tough conversation is not a first-draft situation. That needs to come from you. The difference is whether you are using AI as a starting block or as a replacement for your own thinking.

Question Two: Would a Stranger Get This Right?

Imagine handing the task to a competent stranger who knows nothing about your team, your company, or your history. If a stranger could produce something usable, AI probably can too. Weekly status update formatting? A stranger could do that. Writing a job description based on a role summary? A stranger could handle it.

Now imagine asking that stranger to give feedback to an employee who has been underperforming for six months. Or to decide which of two qualified candidates to promote. Or to navigate a sensitive conversation about workload with someone you know who is dealing with personal issues. The stranger would fail because the task requires specific knowledge that only you have. AI is that stranger.

Question Three: What Happens If It Is Wrong?

Some tasks have low consequences for error. If AI generates a slightly awkward meeting agenda, you fix it in two minutes, and nobody notices. If it produces a status update that misses a nuance, you catch it in review and adjust.

Other tasks carry real weight. A poorly worded performance review can damage trust that took months to build. A tone-deaf email to your team during a stressful period can make people feel unseen. A recommendation that overlooks context can lead to a bad decision. The higher the stakes, the more human involvement the task requires.

Using the Framework

When all three answers point toward AI, use it confidently. First draft, a stranger could do it, low stakes if imperfect. When all three point away, do it yourself. The final product requires your specific knowledge, high stakes. Most tasks fall somewhere in between, and that is where your judgment matters. The framework does not give you a yes or no. It gives you a sense of how much you should rely on AI versus yourself for any given task.

The Over-Reliance Problem

The managers who get the most out of AI are usually the first ones to hit this wall. They start with drafts and summaries, see how much time it saves, and gradually start using it for more. Then more. Then everything. The line between “AI helped me write this” and “AI wrote this” disappears so slowly that most people do not notice it happened.

Skills Atrophy

Writing is thinking. When you draft a performance review from scratch, you are forced to reflect on what that person actually accomplished, where they fell short, and what they need to hear. When you hand that process to AI and just edit the output, you skip the reflection entirely. You are reviewing someone else’s thinking instead of doing your own. Over time, the muscles that make you good at giving feedback get weaker because you stopped exercising them.

Gartner sees this coming. Their 2026 predictions report warns that critical thinking atrophy caused by generative AI use will push 50% of global organizations to require AI-free skills assessments during hiring. The concern is not theoretical. Companies are already starting to test whether candidates can think without the tool.

This applies across the board. The manager who stops writing their own emails loses their voice. The manager who stops building their own agendas loses their sense of what matters. The manager who stops thinking through difficult conversations before having them loses the preparation instinct that keeps those conversations productive. AI did not take these skills away. The manager gave them up voluntarily, one shortcut at a time.

Your Team Notices

People can tell when their manager is phoning it in. They may not know you used AI, but they know when feedback feels generic. They know when an email could have been sent to anyone on the team, and it would read exactly the same. They know when you are going through the motions instead of being present. Trust is built through specificity. When your direct report reads a review that mentions the exact moment they stepped up during a crisis last quarter, they feel seen. When they read a review full of phrases like “demonstrates strong collaboration skills,” they feel processed.

The Dependency Trap

There is also a practical risk. If you cannot write a coherent performance review without AI, what happens when the tool is down? When the company restricts access? When you are in a live conversation and need to deliver feedback on the spot with no time to generate a draft? Over-reliance creates a dependency that makes you less capable, not more. The goal was always to save time on the mechanical work so you could invest more in the human work. If you have simply replaced one with the other, you have not gained anything. You have just outsourced the part of your job that makes you valuable.

The Resistance Problem

On the other end of the spectrum are the managers who will not touch AI at all. They have their reasons. Some think it is a fad. Some think it produces garbage. Some tried it once, got a bad output, and wrote it off entirely. Some just do not want to learn another tool.

The Pride Factor

For experienced managers, there is often an unspoken belief underneath the resistance: I have been doing this for 15 or 20 years without AI, and I do it well. Why would I change? This is understandable. If you have built a career on strong writing, clear thinking, and hard-won management instincts, the idea that a tool could help feels like an insult. It implies that what you do is not that hard. That a machine could approximate it.

But that framing misses the point. AI is not replacing your expertise. It is handling the parts of your job that never required expertise in the first place. You did not spend 20 years learning how to format a weekly status update. You spent 20 years learning how to lead people. The formatting just came with the territory.

The Falling Behind Problem

Management does not exist in a vacuum. The manager down the hall is sending polished project summaries by 9 AM while you are still writing yours at lunch. If every other manager in your organization is producing polished reviews, thorough meeting notes, and well-structured project updates in half the time, your refusal to use the same tools does not make you principled. It makes you slow. Your team does not benefit from you spending two hours on something that could take thirty minutes. They benefit from you spending that extra ninety minutes on the work that actually helps them grow.

The stakes of falling behind are getting higher. Gartner has predicted that by 2026, 20% of organizations will use AI to flatten their structures, eliminating more than half of current middle management positions. The managers most likely to survive that restructuring are the ones who learned to use AI as a force multiplier, not the ones who ignored it.

The Bad First Experience

Most managers who tried AI and quit did one of two things wrong. They either gave it a vague prompt and got useless output, or they expected the first draft to be perfect. AI responds to specificity. “Write a performance review” gives you something generic and forgettable. “Write a performance review for a mid-level engineer who exceeded their Q3 targets but needs to improve cross-team communication, tone should be direct but supportive” gives you something you can actually work with.

The tool did not fail. The input did. That is not a reason to abandon it. It is a reason to learn how to use it properly. The managers who dismiss AI after a single bad experience are making the same mistake as the ones who trust it completely. Both have stopped thinking critically about what the tool can and cannot do.

How to Use AI as a Manager (The Right Way)

The answer is not a formula. There is no percentage split that works for every manager in every situation. But there is a principle that holds up: AI should handle the mechanical work so you can focus on the human work.

What This Looks Like in Practice

On Monday morning, you use AI to generate a first draft of your team update. It takes two minutes instead of twenty. You spend the extra eighteen minutes reviewing it, adding the context only you know, and making it sound like you actually wrote it. That is the balance working.

On Wednesday, you sit down to write feedback for an employee who has been struggling. You do not open ChatGPT. You think about what you have observed, what this person needs to hear, and how to say it in a way that motivates rather than deflates. You write it yourself because this moment matters too much to start with someone else’s words.

On Friday, you need to summarize a dense project report for your leadership team. You paste it into AI, get a clean summary, edit it for accuracy, and send it. Nobody needed you to spend 45 minutes reading and condensing that report manually. The time you saved goes toward preparing for a difficult conversation you have been putting off.

The Ongoing Calibration

This is not something you figure out once and then stop thinking about. Every new task is a small decision about where AI fits and where it does not. Some weeks you will lean on it heavily because the workload demands it. Other weeks you will barely use it because the work in front of you is almost entirely human. The managers who get this right are not the ones who use AI the most or the least. They are the ones who keep asking themselves whether they are using it for the right things. That question never goes away. And it should not.

Conclusion

AI is not going to make you a better manager. You are going to make you a better manager. AI just handles some of the work that was never the hard part to begin with.

The managers who get this wrong fall into one of two traps. They either hand over too much and lose the skills and authenticity that make them effective, or they refuse to engage and spend hours on work that stopped requiring hours a long time ago. Both camps are letting their relationship with a tool define how they manage instead of letting their judgment drive the decision.

Use the framework. Ask the three questions. Pay attention to when AI is helping you think and when it is replacing your thinking. The line will move depending on the task, the stakes, and the person on the other end. Your job is to keep noticing where it is. The St. Louis Fed found that generative AI adoption reached 54.6% of U.S. adults in 2025, outpacing both the personal computer and the internet at the same point in their adoption curves. This is not a niche tool anymore. It is the new baseline. The question is not whether to use it. It is whether you will use it well.

Frequently Asked Questions