Search for “AI for managers” and you’ll find executive courses from MIT and Wharton thought pieces. All fine. None of them look like my Tuesday.
I’m a support manager at a tech company, and at this point AI is involved in almost everything I do as a manager. Not in a futuristic way. In a mundane, “this saved me four hours this week” way.
One thing before we start: this post is about AI applied directly to managing people. Not the automations that reduce ticket volume or deflect demand. I’ve covered that side in AI in customer support. This is about the manager’s chair: developing people, communicating, deciding.
What I Actually Use AI For
The honest list, in rough order of frequency:
Writing in my second language. I’m Brazilian, working in English. I write the substance, AI polishes the delivery.
Messages to the team and to customers. For sensitive announcements or difficult escalations, I draft and then ask AI to poke holes: where could this be misread? It’s good at finding the sentence I was hoping nobody would notice.
Ticket quality analysis. Instead of sampling a handful of tickets and hoping they’re representative, I can review more, faster, with consistent criteria.
Data analysis. Patterns out of support data, trends, the numbers I bring to leadership.
Feedback preparation. Organizing my notes before a difficult 1-on-1. More on the limits of this one later.
Nothing here required training a model or learning to code. These are AI tools for managers in the most boring sense. Text in, better text out. Data in, patterns out.
Three Trainings I Would Never Have Built
Here’s the example I keep coming back to when someone asks how to use AI as a manager for actual people development.
My team needed training. Not generic onboarding material, but targeted programs: customer service quality, social engineering awareness, performance improvement, technical bug investigation. The kind of thing every support manager knows their team needs and almost no support manager builds.
Because building training takes time. A lot of it. Before AI, I would have delegated this to an education or quality team, joined their backlog, and received something generic months later. More honestly: most of it would simply never have happened.
With AI, I built three full training programs, each running about four weeks, in a fraction of the time. The ingredients were mine: I know the subjects, and I know exactly where my team struggles. AI turned that knowledge into structured programs, exercises and materials at a speed that made the whole thing viable.
That’s the shift that matters. The bottleneck was never knowing what the team needed. It was the hours between knowing and delivering. AI removed the bottleneck, and my team’s development stopped depending on another department’s roadmap.
The Experiment That Failed
Now the honest failure.
At some point I got ambitious and dumped thousands of tickets into an LLM, asking for insights. It doesn’t work. With a context that large and varied, the model gets lost. It blends unrelated issues into confident summaries and hallucinates patterns that aren’t there. Asking an LLM to read 5,000 raw tickets is a guaranteed way to get a beautifully written analysis of a support operation that doesn’t exist.
The lesson wasn’t “AI can’t analyze tickets.” It was that structure has to come before intelligence. I’m now building a mixed approach: automation handles the pipeline, each ticket gets categorized individually, everything lands in a database, and the AI queries that structured data instead of swallowing everything at once.
Less magical. Far more real.
The Hours Go Back to People
There’s a quieter benefit underneath all of this.
A lot of what used to eat my weeks was waiting on other teams for small technical things. Now AI works as a tech arm. The clearest example: I built an integration connecting AI directly to Zendesk, so the routine parts of my day, looking up tickets, pulling context, checking patterns, happen in one conversation instead of twenty browser tabs.
The point isn’t the integration. The point is where the recovered hours go: more prepared 1-on-1s, better feedback, training programs that actually get built. AI for managers isn’t about becoming more technical. It’s about buying back time for the part of the job only a human can do.
The Line I Won’t Cross
Here’s where I’m rigid, and I’ve written about why in Human Skills in the Age of AI: I refuse to delegate decisions about humans to AI.
AI does not decide who on my team is performing and who isn’t. It can organize my notes and help me structure a hard conversation. The judgment, the knowledge that this dip has a story behind it, stays with me.
Same rule for feedback. AI helps me prepare it, but I never deliver it without my own review on every line. The moment your team suspects your feedback is machine-generated, your credibility is gone. And they’d be right to doubt you.
The pattern, if you want one: AI handles volume, humans handle judgment.
Bringing the Team Along
My team knows I use AI for all of this. I talk about it openly, including the failures, and I’ve been nurturing the same culture in them. The company helped by giving us AI subscriptions connected to our actual work tools, so nobody has to ask permission to start.
Some ran with it. Others are slower, and that’s fine. The goal isn’t everyone building integrations by Friday. It’s nobody feeling like AI is a thing that happens to them instead of a thing they use.
That’s AI for managers after a couple of years of daily use. Not a course. Not a transformation roadmap. A capable assistant, a tech arm I didn’t have before, and a set of decisions I deliberately keep human.
What’s the first thing you’d build for your team if the hours weren’t the problem? That’s probably your starting point.