AI: Lessons from a Vibe Manager

Everyone’s a vibe coder now. Or a vibe engineer. Or an agentic engineer if you think vibe coding is beneath you.

I am a Vibe Manager. While the name sounds silly, the idea behind it is real.

I built a technology services business from scratch and ran it for 25 years. Along the way, I built deep expertise in several technical areas, but really, I became a good manager.

Most people treat AI like a vending machine. They punch E4, expecting Sour Cream & Onion, and get Veggie Stix. The next day, they punch E4 again and wonder why they still don’t have Lay’s.

I was guilty of this, too. Then I realized, I don’t have to become an agentic engineer to get the most from AI. I need to manage my AI team the same way I manage people: with clear roles, regular feedback, and real accountability.

I treat each new agent like a new employee. I explain what the business does, what good looks like, and when to stop and ask for help. For one agent, I wrote an onboarding doc that started with “You’re not a content mill. You’re an inbound content strategist designed to help fill content gaps in specific silos. You will soon learn what content we have and what we need.” When it drifted, I narrowed the scope and corrected early. I did not say, “You idiot”; I explained what I wanted and why. Within a few sessions, it was producing research-backed, draft-ready work with minimal direction.

The AI conversation is driven by two camps: prompters and builders.

The prompt camp trades tricks like recipes. YouTube is rife with “This One Prompt will Change How You Use AI Forever”; heck, it was even the title of a TEDx talk. Search Indeed, and you can even get a “Prompt Engineer” job making $100K a year.

Then you have the builders. Professional developers building products for enterprise use. The non-developer enthusiasts who realized that with a little code and some no-code tools, you can automate a lot of stuff and make some pretty cool things.

There’s even a guy on YouTube with 43,000 subscribers building an entire platform around vibe coding: Merch and everything. I don’t even want to talk about it.

Peter Steinberger — the engineer behind OpenClaw — told Lex Fridman, “I think not a lot of people ever considered the way the agent sees the world.” Lex called it empathy.

That’s management. Every new hire walks into a business they don’t understand, with context they don’t have. A good manager closes that gap — scopes the role, provides context, and gives feedback early.

That’s not an engineering skill. It’s a people skill.

I’ve been helping my team for twenty-five years. I only recently realized it was an AI skill.

I have a client using AI for SEO. The AI is doing a fine job — EEAT, on-page optimization, JSON-LD schema, all solid. But the business problem isn’t SEO. It’s conversion and buyer engagement. The AI is doing exactly what it was told to do. It just wasn’t told the right thing.

You wouldn’t blame an employee for doing the job you assigned them.

AI engineers see this too, just through a different lens. Read any guide on multi-agent systems. Engineers talk about “role definition,” “scoped responsibilities,” “delegation patterns,” and “coordination.”

That’s not engineering. It’s team building.

Anthropic even called one of their latest features: Agent Teams.

Engineers reinvented management from first principles and wrapped it in technical language because that’s their world.

The management world is simpler.

And when somebody keeps making the same mistake, you don’t fire them. You figure out why. Maybe you provided bad instructions. Maybe you keep punching E4 and expecting Sour Cream & Onion.

Good management develops people. Walmart’s new CEO, John Furner, started as an hourly associate in a garden center. You give people more responsibility as they earn trust. Agents work the same way.

The people best positioned to get value from AI aren’t engineers — they’re the ones who’ve been managing teams for years and think AI isn’t for them because they’re “not technical.”

If you’re a manager trying to adopt AI for your business, you are closer than you think.

Much closer than the developer who’s never managed anyone.

This piece was written with the help of AI agents — managed, not prompted.