Vision First, Agents Second
If you can't articulate the vision clearly enough for a human team, what makes you think you'll be able to orchestrate a fleet of AI agents?
Listen to under 2-min AI recap •TL;DR ↓
This article contains affiliate links. If you sign up through them, I may earn a commission at no extra cost to you.
Walking in the Park, Shipping Software
I've been walking in the park lately. Occasionally I pull out my phone, answer my AI agent's questions, instruct it on what to work on next. Then the screen goes off, back in the pocket, back to the trees. I used Claude's mobile app for this.
This is a significantly larger chunk of work than I've delegated before. At the end of the day I'll review the result myself before shipping — just to see how well my instructions landed and what came back. Not diving into the technicals yet. It's still an experiment.
But it's a telling one.
The Pattern Repeats Itself
Not so long ago, businesses were onboarding to the internet. Building representative websites, trying to be visible online. Many businesses to this day just operate. They operate to make money. Some have a vision. Some are in survival mode.
I remember building those websites 15, 20 years ago. I remember businesses struggling to provide content. Many wanted pages like "Vision" and "Mission" — but some couldn't actually define those things, despite wanting to represent themselves well.
The same pattern shows up in software engineering. Developers keep building a product without necessarily understanding how they're going to reach the end result. Moving forward in an erratic way, but somehow getting there.
Not all of them, though. Engineers identify a challenge, define it, analyse it, design the solution. They hold a vision of the end goal and a clear picture of how to get there.
The Moral Is the Same, Just in the AI Era
Connect both topics and the lesson carries over directly.
If we know what we're doing — if we hold the full-scope knowledge, the picture, the vision of every process and every step (which could be owned by many people) — only then can we build a predictable, well-designed, maintainable solution.
That was true for websites. It was true for software products. And it's true now with AI.
But zoom out. Think about this from a broader perspective.
A massive number of AI agents working together. The concept has been in the air for a while — entirely AI-driven companies. Today they're more realistic than ever before.
I have that picture in my head. Not as speculation. As something I'm starting to taste on my afternoon walks.
If you can't articulate the vision clearly enough for a human team, what makes you think you'll be able to orchestrate a fleet of agents?
TL;DR
- The ability to clearly define vision, process, and end goals has always separated great builders from everyone else — AI changes nothing about that.
- Many businesses couldn't even articulate a mission statement for a simple website; now they're expected to orchestrate autonomous agents.
- Engineers who hold a clear picture of the full scope and every step are the ones who produce predictable, maintainable results — with or without AI.
- Delegating real work to AI agents is already happening today, from a phone in a park, and it works when the instructions are precise.
- Entirely AI-driven companies aren't speculation anymore — they're an inevitable extension of this delegation pattern.
If you can't articulate the vision clearly enough for humans, you'll never be able to orchestrate a fleet of agents.
Geddy
Senior Web Engineer / Lead
Engineering leadership • AI innovation • Product thinking. 20+ years of web engineering, from independent contractor to engineering leader. Passionate about developer experience and product engineering.