⚙ Hi {{first_name_tally|Operator}},

My Amazon account just got flagged for adding 97 new shipping addresses shipping out your thank you books for completing our annual survey.

Almost 80% are using ChatGPT at work, but the most popular automation tool (Zapier) was only at 14% usage.

AI means operators are in higher demand than ever. But they have to be able to build a system beyond themselves. That can be with people or machines, so long as it produces results.

I run 3 businesses, including The Bottleneck. That means 6-8 meetings a day—some on Zoom, some in person.

I used to spend half my time writing up what just happened instead of doing the actual work.

Now I don't take notes at all.

Circleback records everything. On Zoom, the overlay sits quietly in the corner—no awkward "AI bot has joined" notification. Meeting ends, notes hit my inbox with action items already assigned.

But here's what sold me: the in-person stuff. I pull out my phone, hit record, and forget about it. Twenty minutes later I have a full transcript and know exactly what was discussed.

Most AI note-takers only work on video calls. This one works where I actually work.

Every scaling company hits the same wall. Not enough people who can drive projects end-to-end. Too much work. And the only answer anyone offers is "hire more."

Keith Rabois called this the barrels-and-ammunition problem. Barrels are the people who can take an idea from concept to launch while bringing others along. They're your constraint. Everyone else is ammunition: talented executors who need someone to aim them.

For twenty years, the playbook wrote itself. Find barrels. Hire ammunition. Ship more. If you wanted to scale outbound, you hired SDRs. If you wanted more features shipped, you hired engineers. Headcount was the answer.

Then I met someone who broke the model.

Andrew went dark on a Tuesday. 14 hour work days, a breakup that went sideways, and a health scare meant he needed a few months reset. When your best sales guy vanishes, deals die, meetings cancel, and your funnel leaks until you backfill. That's usually the kind of disappearance that's supposed to crater your pipeline.

Except ours didn't.

Heyreach sequences kept firing. LinkedIn automation ran on autopilot. Meetings booked themselves, two or three a week, because Andrew had built the system before he left. Lead list loaded, responses mapped, my assistant just pressed send.

We rented his account and let the machine hum. Copied it to two more accounts. Same playbook, three times the output, zero new headcount.

Andrew figured out something most operators never do: the new barrel doesn't hire ammunition. They manufacture it.

The person who's great at managing ten people is rarely the person who can build the system that replaces them. Managing humans is one skill. Designing workflows is another.

Andrew never managed anyone. He was an IC through and through. But he could look at his own work and ask one question: what here is actually hard? The parts that just looked hard because nobody had mapped them, he turned into systems.

Not someone who scales through people. Someone who scales through process.

Before Andrew automated anything, he mapped everything.

Every script became a flowchart. Every objection became a branch. Every follow-up sequence became a diagram with clear inputs and outputs. He was writing the spec for his own replacement.

The test is simple: can you document it with flowcharts and checklists alone?

If the work follows a predictable path, you can map it on a whiteboard. Inputs, decision points, outputs. If you can do that, you don't need a person. You might not even need AI.

LinkedIn outreach, chat responses, follow-up cadences. All flowchartable.

Most roles bundle the flowchartable 80% with the judgment-required 20%. Companies hire full humans to do work that's mostly deterministic. Andrew's system books 8-12 qualified meetings per month. An SDR doing that volume costs $65K loaded. His automation runs on $200/month in Heyreach and Clay.

That's not efficiency. That's a different category of leverage.

Not everything needs AI. Most things don't.

Most work sits on a spectrum. At the bottom: deterministic systems. Cron jobs, spreadsheet formulas, if/else scripts. They run at 3 AM exactly the same way they run at 3 PM. No moods. No context drift. No hallucinations. Boring and bulletproof.

One level up: workflow automation. Zapier, Make, n8n. Still deterministic, still reliable, but handles complexity across systems. Andrew's Heyreach sequences sit here. Multi-step flows with branches, but no judgment required.

Then: AI with guardrails. LLMs with human review. Structured outputs. Human starts it, AI handles the middle, human approves the end. You're paying for the AI's pattern-matching, but you're not trusting it blindly.

At the top: humans. Expensive. Flexible. Required for novel situations, relationship building, taste, direction, strategy. The stuff that can't be flowcharted.

Salesforce learned this the hard way. They spent millions building Agentforce, AI agents handling enterprise workflows end-to-end. The agents kept losing context, drifting from goals, forgetting what they were supposed to do. Their fix? They built an "Agent Graph" system that added deterministic logic back into the workflow. Flowcharts and checklists. The same if/else structure a Zapier flow would've provided from day one. They spent millions rediscovering why McDonald's uses a timer for fryers.

Reach for scripts before AI. AI before humans. The simplest tool that works is usually the right one.

I have a great EA. But expecting her to manually track recurring tasks across five different systems is asking her to do deterministic work badly. She was worried about being automated away when I first asked her to flowchart how she handles emails. I asked for trust. We carried on.

The first week we automated the predictable 70%, shuffling tasks and emails between tools, I could hear the stress melt from her voice. She stopped drowning in administration and started catching problems before they grew. That's not automation replacing her. That's automation promoting her.

I've watched companies hire ammunition when what they needed was someone to flowchart the existing work. Good operators quit because they were used as ammunition when they should have been barrels.

Watch what your team asks for when they hit capacity.

The old barrel says "I need another person." Stable, but slower. Fine for leading leaders. Not fine for scaling output.

The new barrel says "let me map this out and see what's actually hard." That's the difference. Find ten more of them.

Andrew's flowcharts became the spec. His departure was the stress test. The system passed.

The barrels-and-ammunition framework got the constraint right. Organizations ship at the speed of their barrels. That hasn't changed.

What changed is how barrels create leverage. It's not just hiring anymore. It's flowcharts that execute, scripts that run, workflows that fire at 3 AM while everyone's asleep.

The companies eating your lunch aren't outworking you. They're not out-hiring you. They're not even out-AI-ing you.

They're running the same playbook with one-third the headcount because their barrels know the difference between a problem that needs a person, a problem that needs a prompt, and a problem that needs a for-loop.

Rabois was right about the constraint. But the barrels that win in 2025 aren't the ones who rally teams around problems. They're the ones who look at their own job and ask: which parts of this should outlive me?