Two Claudes, One Contract: The Changing Face of Decisionmaking
Kraftisms #32 - A decision is now two calls, not one. The context call. And the judgment call.
Quote(s) of the Week
“If I’d asked people what they wanted, they’d have said faster horses.” -Henry Ford
“Every battle is won before it is fought.” -Sun Tzu
Kraftism of the Week: The Lists That Didn’t Overlap
A few weeks back, on a Tuesday, an executive I work with forwarded me a contract to review. He also included a list of “problem” items to push back on. Eleven of them.
The exec had run the partner’s redline through Claude. Claude looked at the contract cold - no context about the partner relationship, no material record of prior pushbacks we’d accepted from over a hundred similar negotiations, no sense of where our usual fallback positions live - and gave him eleven red flags.
I’m not the lawyer here. I’m doing biz dev for the company - which on this engagement means redlining standard agreements. And, with each new contract, I’ve been teaching Claude the actual contract patterns. What we sign. What we don’t. What we trade for what. Where we hold a line. Where we don’t bother.
So when the exec forwarded me his eleven-item list, I ran the same contract through my Claude. Same tool. Different context.
My list was five items - and the lists didn’t even fully overlap.
Claude didn’t get smarter. Claude was the same. The five items were the five things that actually mattered for this partner, this deal, this stage. The other items from the original list were either standard language for this counterparty or things we’d handled creatively. Because that’s hard for Claude. I know one line can invalidate an entire exhibit - but Claude still wants the edits inside it, not catching the creative way of handling it. I then took in those five and implemented four of them - and one of my own that even Claude didn’t catch. I made a judgment call.
When I sent the exec the edits, he wasn’t surprised. He’d run his pass to see potential issues. He trusted that I had the context for the actual redlines. He told me that’s the kind of person he looks for now. The one who shows up with the context already loaded.
That was the moment. I’d been doing the upfront work for weeks. Hours of it. I just thought I was doing the fractional work I had been brought on board for. He spotted what I was actually doing - and named it as the thing - before I did.
I’ve been on the Listening Tour for a little over two years now. Five-plus meetings a week. Something I’d been hearing for months suddenly fit a shape.
Applying the Kraftism: Context, Then Call
So.
Every wave of workplace automation in the last forty years has done two things at once. It compressed the labor in the middle. And it redefined what counted as the work-product on the ends. AI is doing both. And it’s doing one more thing the prior waves didn’t.
That exec’s framing has been showing up in my Listening Tour conversations all year. The shape of the work is changing, and the people who notice early are the ones running the play differently next quarter.
Format, Layout, Production
Typewriter to word processor. Format work disappeared. Margins, tabs, fonts, line breaks - absorbed by the tool. What survived was the writing. The writer’s job didn’t get easier. It got purer. The hard part rose to the top because it was the only thing left.
Manual layout to desktop publishing. Layout work disappeared. Cutting, waxing, photo-stat reduction, mechanical line spacing - absorbed by PageMaker, then Quark, then InDesign. What survived was layout judgment. Not the act of laying out. The call about what should go where. Designers got cut. The ones who stayed got paid more.
Film crew to phone-plus-social. Production work disappeared. Lighting, multi-cam, color grading, sound mixing - absorbed by an iPhone and a free app. What survived was the choice about what the video said. Brands that spent twenty thousand on a thirty-second spot now ship five spots a week and the only hard part is deciding what each one says.
Three waves. One pattern. The labor compressed. The judgment survived. The people who recognized the shift early - and reorganized around the part that survived - won.
Both Bookends, One Decision
AI compresses the labor too. That part is familiar. But it adds a layer the prior waves did not require, and it changes what “making a decision” means.
The word processor didn’t need your audience. Spell check didn’t need your tone. CapCut didn’t need your brand voice. AI does.
The decision is no longer just the call at the end. The eleven-item list and the five-item list both came from the same model. The difference was made before either list existed - in the hours of context-setting upfront. The decision was already half-made by the time the model saw the contract.
In the prior waves, the human made one decision: what to write, what to design, what to film. The labor compressed; the decision stayed at the back. AI changes that. The decision is two-sided now. Front: what context the model gets. Back: what to do with what comes back. They aren’t two decisions. They’re one decision with two ends. Pull either end out and the decision falls apart.
Context is now decision-grade input. That makes the wires that move it - from system to system, from publisher to buyer, from agent to agent - infrastructure, no longer a nice-to-have.
Look at This Week
Tom Triscari pointed out this week that The Trade Desk just joined the Google-Meta club on reporting revenue per employee. Meta and Google have hit headcount maturity. The most-valuable adtech firms in the world are not adding people. They are keeping the operators, the decisionmakers, and letting agents do the labor in between.
Publishers who treat their first-party data as portable context - what works on their audience, what they’ve learned about who reads what when - are setting up the same way. The first-party data isn’t the deliverable. It’s the training the buyer’s model needs before any buy can be smart.
Stagwell launched an SMB agent cloud. Hershey is putting agents on its measurement stack. Omnicom is routinizing agentic buys. Every announcement reads as an efficiency story. Every announcement is actually a quiet bet that the value has moved to the framing of the work and the judgment about both its input and its output - and the labor in between is becoming a commodity.
To Wrap It All Up
Three questions, depending on which seat you’re in.
If you run an agency or a planning team: Are you measuring your people on the deck they ship, or on the context they set and the calls they make on what comes back?
If you sell publisher inventory or first-party data: When you talk with a buyer, are you handing them a deck they can port, or context only you can give - and then standing alongside them while they judge what their model does with it?
If you lead anything: When you hire next, are you hiring for the ability to generate a work product, or for the ability to make smart decisions on both bookends?
Decisions don’t sit at one end anymore. They sit at both. Show up with the context call. Stay for the judgment call.
Andrew.



