The only move that matters.
A quick note on why I'm up here.
— 2017, AppSumo —
We're learning together. This talk is a contribution to a conversation you're already having.
There is nothing so useless as doing efficiently
that which should not be done at all.
Peter Drucker · 1963(63 years ago)
Each era pulled away from the one before it.
This one will pull away the fastest.
A customer signal travels four hops before it reaches you. Three people already decided what you needed to know. The bottleneck is approvals.
Every decision flows through one person. It scales to ~30 people. Then it breaks. The bottleneck isn't bureaucracy anymore. It's you.
Intelligence sits at the center. Every signal feeds in directly. No relay. No decay. Every signal makes the system smarter for the next one.
Most of you are implementing AI.
Most of you are doing it wrong.
Founders installing Open Claw
instead of solving the chokepoint in their business.
A flywheel is the loop in your business that compounds value with every turn.
Each cycle makes the next one easier. Each customer makes the next acquisition cheaper. Each insight makes the next decision faster.
Find yours. If you can't draw it on one page, that's the work.
Drop the question in at the top. Keep asking "which side has more friction?" until you hit the bottom.
| Company | Flywheel | Bottleneck | AI at the Bottleneck |
|---|---|---|---|
| Sales | Call → close or lose → learn → next call | Real-time learning across reps | Every call transcribed and scored. Winning phrases surface in real time. Scripts update themselves. |
| Support | Ticket → resolve → learn → resolve faster | Institutional memory across tickets | Macros update themselves from every resolution. Knowledge stops living in people who quit. |
| GovTech | Submit → reject → learn → resubmit | Learning from each rejection | Model captures every failed clause and reviewer. Approval rates climb. Cycle times compress. |
| Agency | Campaign → result → apply → next campaign | Cross-client pattern recognition | Central model surfaces relevant past wins per brief. Junior work matches senior output. |
Same move. Four times. One principle in four clothings.
These aren't outliers. They're the early signal.
One named human on the hook. Not "the team." Not "IT."
A single number it exists to move. If you can't state it, kill the agent.
The threshold at which you turn it off. Defined before it's turned on.
If your biggest competitor
copied your entire tech stack tomorrow...
would they get the same results?
You don't have a moat. You have a subscription. Anthropic and OpenAI own your edge.
Your data, your customers, your history — none of it transfers. The loop is yours. That's the moat.
Here's the part
no one says from the stage.
"Claude-pilling" the company means mandating daily AI use as a condition of doing the work. The ones who pick it up will 10x their output in 90 days. The ones who can't — or won't — become dead weight in a lean org. Companies are quietly using AI adoption as the silent criterion for who stays and who goes.
The divergence is already measurable.
Companies that architected are compounding. Everyone else is getting cut.
Everything else can and should be delegated.
Everything else can and should be delegated.
Where are we going?
AI can improve the map.
It cannot choose your mountain.
As intelligence becomes abundant, conviction becomes scarce. You set the meaning. The machine handles the math.
Who are we building with?
One high-agency operator with AI fluency
outperforms ten average ones.
Stop hiring for headcount. Start hiring for multiplier. Architects design. Assemblers refine. Amplifiers distribute. Everyone else is overhead.
What are we deploying?
Cash, tech, attention, conviction —
every finite resource you control.
Capital. The tech stack. Your team's calories. Your own focus. And the oldest one: never running out of cash. The CEO doesn't manage assets one at a time. They allocate them as one portfolio.
Point AI at the
chokepoint.
Everything else is noise.
Remember Client A?
Headcount 400 → 40. Revenue doubled.
They didn't wait until they had the perfect AI strategy.
They didn't wait for the tools to mature. They didn't wait for permission.
Start before you feel ready.