Jerk Report #3 - Your 30%

A major tech company, Block, laid off half its workforce last week. Nearly 4,000 people.¹

The stock went up 24%.

The CEO was blunt: AI tools have fundamentally changed what it means to run a company. He predicted most companies would reach the same conclusion within a year.

The market's message: fewer humans, more value.

If that sentence made your stomach tighten, good. That's signal. Let's read it — and then let's find where the agency actually lives.

Position

The standard advice right now is: become an AI orchestrator. Don't write the report — direct the AI that writes it. Don't do the analysis — prompt the system that does.

That's where most of the conversation is. Position.

Velocity

Software developers were the first to live this. Vibe coding — building by describing what you want instead of writing the code — went from novelty to normal in eighteen months. I wrote about this last week: it gets you to 70%. Fast and cheap. The last 30% is where businesses live or die.

We're already seeing what happens when people skip the 30%. Apps that crash in production. Workflows that lose customer data. Security holes nobody tested for. One client put it perfectly: "I kept painting myself into corners and trying to back out."

The 70% looks like a finish line. It's a starting point — a prototype that reveals what you didn't know you needed. That matters because this pattern is spreading beyond software into legal, marketing, finance, operations. The crashes will follow when people in those fields treat the 70% as done, too.

Acceleration

Now the orchestration itself is being automated. AI agents that chain prompts, test output, and iterate — without a human needing to review each step. The thing people just learned to do is already being done by AI. Of course, you should keep a human in the loop still, despite what AI companies might say.

The role isn't just changing. The rate of change is speeding up.

Jerk

Each new AI model moves the line between what humans handle and what AI handles. Not gradually — in jumps. Each capability release triggers another round of "wait, it can do that now?"

The rate at which organizations are redefining what they need people for is itself accelerating. Discontinuously. That's jerk.

"Become an orchestrator" might be obsolete before you finish implementing it.

Where the Agency Lives

That sounds bleak. It isn't.

The people who are thriving are not the best prompters. They're the ones who know what the work is actually for. They can tell when AI output is plausible but wrong — because they have the judgment to evaluate it.

The human doesn't go away. The humans' roles change. And the people who know where their judgment lives have the most agency right now.

Let me make that concrete.

What This Looks Like in a Bid Process

Say your team produces estimates or proposals. Watch how the 70/30 line moved in one year.

A year ago, the whole bid was human work. Site visit. Measurements. Material lookup. Labor hours. Allowances for conditions you've seen before. Write-up. Format. Send. Nobody called any of that "judgment." They called it "doing the bid."

Six months ago, AI could draft the write-up from your notes. Maybe pull standard material pricing. You still did the site visit, the measurements, and the thinking. AI handled maybe 30%. The line: 70% human, 30% AI.

Today, AI can take photos from the site visit, extract measurements, pull current pricing, calculate labor from historical data, and generate a formatted proposal. It gets you to 70% of a finished bid.

The human is now in the other 30%: Does this price reflect what I actually saw on site? Did it miss the drainage issue that adds two days? Is this the kind of client where we adjust margins, and why? Does this bid make us money or just keep us busy?

That line moved from 30% AI to 70% AI in twelve months. The next model release will move it again.

Three Things You Can Do This Week

These use the bid process as an example. Substitute whatever process touches your revenue.

1. Name the judgment calls in one process.

Walk through your last bid. Where did someone make a decision that required experience, not just data? Maybe it was adjusting labor hours because you know that building's parking situation adds half a day. Maybe it was shading the margin because this client leads to three more. Maybe it was catching that the AI pulled pricing for the wrong material grade.

List those decisions. Write them down. Those are your 30% — the part that survives. If you can't name them, that's a signal. It means the judgment is invisible, which means it's at risk.

2. Map how the line moved.

What could AI do in your bid process a year ago? Six months ago? Today? Is the line moving steadily or in jumps? If it jumped when a new tool launched, that's jerk — not smooth change but discontinuous shifts. Jerk requires a different response than gradual automation. You don't just learn one new tool. You build the capacity to keep adapting as the tools keep changing.

3. Build your value on the 30%, not the 70%.

The temptation is to get faster at the 70% — better prompts, more automation, tighter templates. Fine. But that's a race you eventually lose, because AI gets faster at the 70% too. The durable investment is deepening the judgment AI can't replicate. The site visit instinct. The client relationship. The pattern recognition from a thousand jobs that tells you when a number doesn't smell right.

The company that laid off 4,000 people still employs 4,400. The signal isn't in who got cut. It's in what those people do that the others didn't.

Know what yours is.

If reading the derivatives of your own business sounds useful, as in figuring out your 30%, I do that. Half-day intensive. https://www.designrosetta.com/landscape-analysis-introductory-conversation

Rose Thun writes the Jerk Report every Monday. Subscribe at designrosetta.com/jerkreport

¹ Block, the parent company of Square and Cash App, laid off more than 4,000 employees. CEO Jack Dorsey's letter to shareholders stated that "AI tools have fundamentally changed what it means to build and run a company" and predicted "within the next year, the majority of companies will reach the same conclusion and make similar structural changes."