AI
The Product Managers Who Delegate Best Get the Most Out of AI
At 6 a.m., my house is a small battlefield. Someone is hunting for socks. Someone hasn't showered. Someone suddenly remembers homework.
If I did everything myself, nobody would leave on time. So I learned one thing early: handing work to someone else is a skill, and skills can be trained.
Lately I noticed that the same skill is the one that matters most when working with AI.
I'm Brian Arfi Faridhi. This post is about why product managers who are used to delegating usually get results from AI fastest. And why people who struggle to delegate to humans usually hit a wall with AI too.
Delegating at home is the same as delegating at work
My first lesson as a father of several kids was simple. "Please clean your room" is not delegation. It's a hope.
My kid will clean it their own way. The toys go into the closet, but the closet is stuffed so full it won't close. Technically done. The result still annoys me.
What makes it work is clarity. What the end result looks like. Where the boundaries are. When I'll check. If those three things are clear, I don't need to stand behind anyone.
Work is exactly the same. I've shipped products for almost twenty years at Tokopedia, Hijra, and Flip, plus three startups of my own. At Tokopedia I led a team of 3 PMs, 25 engineers, and 5 designers. At Hijra I guided 5 PMs across several product flows. Today I lead four product teams.
I could never do all of that myself. So what I trained over the years wasn't coding or design. It was giving tasks with enough context, a clear definition of done, and a sensible check-in schedule.
AI is a very fast team member who can't read your mind
When I started using AI seriously, it surprised me how much it felt like a new hire. Smart, but without context.
It's diligent. It's fast. It never complains. But if your instructions are half-finished, it fills the gaps with guesses. And it guesses with total confidence.
People who delegate well already know this. They give background without thinking about it. They describe what good looks like. They say what not to touch. And they aren't surprised when the first draft is off, because that's normal. You give feedback, and round two is usually much cleaner.
People who are used to doing everything themselves often get frustrated. They type one short sentence, the result misses, and they conclude "this AI isn't smart." But if you gave a new hire an instruction that short, the result would miss too.
I've seen this pattern repeat inside my own teams. The people who get good fastest are not the most technical. They are the most patient at explaining.
What I delegate to AI today
I now use AI at work and on side projects, with the same pattern I use to manage people.
At work, I helped push Gen-AI support automation from 0% to 70% of roughly 10,000 tickets a month. I also built FINA, an agent-based customer support orchestrator, as a proof of concept. I got the cost down to about $0.004 per conversation turn. During review, my team found a gap that could expose one user's data to another, before it reached production. That's why delegated work still needs checking.
Outside work, I built an 8-channel content distribution system that I run alone. One long video gets cut into clips and spread across YouTube, Shorts, Reels, Stories, WhatsApp, LinkedIn, and Threads. I review, and I say go.
I also built the Applied-AI Certification and run AI Circle, a community for working professionals who want to learn this way of working.
Read that back and the pattern is the same as managing kids at home. I explain the end result. I set the boundaries. I check at the points that matter. The rest, they run.
Old habit, new tool
One thing I want to be fair about. Saving companies money is an old habit of mine, long before AI was trendy. It adds up to $4M+ per year, from many different initiatives: process improvements, cost optimization, and more efficient product decisions. Automation was only one of them.
In those days, leverage like that needed a title, a team of engineers, and expensive systems. Now AI is the sharpest tool for the same habit. And that habit can be trained in everyone on a team, not just the people with seniority.
This is where delegation skill decides things. Everyone can get the same tools. What differs is the ability to break a problem down, give context, and decide when something counts as done.
I'm not saying delegating to AI is easy. Sometimes I trust too much and check too late. It happens with people too. Trust without checking isn't delegation. It's resignation.
Start small
If you're a PM or team lead, you already have the foundation. Pick one repeating task you'd normally hand to a team member. Write the instructions exactly as you would for someone new: the context, the end result, the limits. Give it to the AI, then give feedback the way you would to a person.
You don't need to understand every tool first. Your habit of managing people already gets you halfway.
If you want to learn alongside others heading in the same direction, join AI Circle. If you lead a team and want this habit to spread to everyone, see the corporate page. Take whichever fits you.
And before all that, this morning when you tell a kid to find the missing sock, notice how you give the instruction. It's the same practice.