The Bottleneck Moved
1. The Setup
I’ve watched teams grind through review cycles, knowing something was off but not being able to name it. Creative round after creative round. Feedback that contradicts last week’s feedback. Beautiful work getting scrapped not because it’s bad, but because… something shifted. Then I read a piece by Nate B. Jones (AI expert I follow) and one line stopped me: “The meeting takes longer than building the prototype.” His thesis: execution used to be expensive, so we built rituals to protect it. Approval gates. Alignment meetings. Review rounds. Now AI made execution cheap, and those same rituals became the bottleneck. The bottleneck moved. And nobody updated the map.
2. My Approach + AI Role
I took Nate’s idea into a conversation with AI to see if it held up against my own experience. His frame is developer-centric (“ship rough, test fast”). Agency world is different. Polish IS the product. You can’t tell a creative team to present half-baked work. So I used AI to stress-test it. Where does this hold? Where does it break? What am I not seeing?
3. What Actually Happened
The thesis held. But it needed translation. Look, agencies are basically deck factories. (I say this with love and self-awareness.) We build decks to think, decks to sell, decks to align, decks to present the deck we’re going to present later. It’s how the work moves. And there’s a reason for that. Agencies that skip the strategic and conceptual layers to compete on execution speed? Their work doesn’t land. Clients come back to us after getting burned by the “fast and cheap” shop that jumped straight to making things without doing the thinking. The layers exist because they work. The problem isn’t the layers. It’s when you’re still validating direction but everyone thinks you’re executing. What I noticed: teams weren’t over-polishing because they loved craft. They were polishing because they’d been burned before. When work gets killed three times, you start perfecting everything. That way if direction shifts again, at least no one can say the work wasn’t good. It’s a rational response to an irrational system. I started thinking about where investment should actually go. At the strategy layer, the thinking should be sharp. The articulation matters. But the visuals? Keep them rough. Sprinkle in enough to bring it to life, but don’t make it precious. A polished deck can mistakenly signal “this is decided” when you’re still figuring it out. Yes, clients need something they can flip to their stakeholders, so a certain polish is expected. But save the heavy craft investment for when direction has been validated and there’s something stable to build on. What I keep seeing: teams building execution-level decks while they’re still answering strategy-level questions. Burning hours on presentations for directions that haven’t been validated yet.
4. The Real Insight
AI doesn’t replace judgment. It reveals where judgment was missing all along. When execution gets cheap, you don’t just get efficiency gains. You expose all the clarity work that wasn’t actually happening. Teams weren’t slow because execution was hard. They were slow because direction was unclear, and expensive execution forced everyone to pause and pretend to align before committing. Now that pause is gone. The lack of clarity is visible. The reframe I’m using:
- Not “go faster.” – “Reclaim headspace from churn.”
- Not “iterate rougher.” – “Stop burning craft hours on directions that haven’t landed yet.”
- Not “be more efficient.” – “Give craft a stable foundation before asking for excellence.” Teams don’t resent speed. They resent feeling constantly reactive.
5. Try This If…
If review cycles keep spinning and you can’t name why: Ask what layer you’re actually iterating on. If feedback keeps contradicting itself, you might be iterating on strategy while everyone thinks they’re iterating on creative. If you’re leading a team through AI adoption: Don’t lead with faster. Lead with “less wasted effort.” That’s a message people can actually hear.
6. Systems Lens
A tension I’m still sitting with: faster iteration can accelerate dysfunction. If the client can’t get internal alignment, your 10x speed just burns through cycles faster. Your speed becomes their stress. And there’s a change management layer here. “Do it faster with AI” can land as “your thinking is replaceable.” That’s not a process problem. That’s a trust problem. The thesis I’m holding: the bottleneck moved. Clarity, context, trust. That’s what’s scarce now. I’m focusing there.
What are you noticing in your own teams? I’m curious if “polish before the foundation is stable” shows up differently in other contexts.