Chapter 17: Redesigning the Work

Chapter 17 · Part Four: The Machine That Makes Average

People say “AI is going to take my job,” and the sentence hides four different things inside one word.

PEOPLE. Judgment, relationships, accountability.

OUTCOMES. What the business actually needs.

WORK. The activities that produce it.

TASKS. Research, formatting, synthesis, admin.

AI is eating the task layer. It’s starting on parts of the work layer. It isn’t eating the outcome, and it isn’t eating who owns the outcome.

A job was never a list of tasks. It was a promise to deliver an outcome. The tasks were just how we used to do it.

That distinction sounds comforting, and it’s meant to be, but it has a sharp edge. If your job is defined by its tasks, then AI really is coming for it, because the tasks are exactly what’s getting cheaper. The only durable version of a job is the one defined by an outcome somebody owns.

Which brings me to the four most expensive words in marketing.


“That’s not my job”

Listen to how each function in a typical marketing team describes being done.

  1. Demand gen: we generated the leads and put them in the CRM.
  2. Ops: we ran the process.
  3. Brand: we made the campaign.
  4. Content: we published the asset.
  5. Product marketing: we shipped the messaging and the enablement.
  6. Events: we delivered the booth on budget.

Every one of those is true. Not one of them is an outcome.

“That’s not my job” is the comfort instinct in its purest form. It draws a line around the tasks you were assigned and declares everything past the line somebody else’s problem. It feels like focus. It’s actually the reason nobody owns the thing that matters, because the thing that matters always sits on a line between two functions.

Take the most common example. Marketing produces a lead, hands it over, and is done. Sales takes the first meeting. Whether that first meeting is any good, whether the buyer leaves it thinking their time was well spent, belongs to nobody. Marketing says it’s a sales problem. Sales says marketing sent them junk.

Here’s the trap I want to flag, because I nearly fell into it writing the talk. The obvious redesign is to say the outcome is “pipeline.” It isn’t. Go back to Chapter 6 and run the test: you can generate a great deal of pipeline while the business gets worse. Pipeline is an activity with a dollar sign on it.

The outcome is a first meeting worth the buyer’s time. That’s something that happens in somebody else’s head, in a room you’re not in, and you can’t achieve it while the business gets worse.

Now notice what changed. Not the tooling. The definition of done.


The better first meeting, built

Here’s what that redesign looks like as a real system rather than a slide. It’s also in Appendix F, ready to run.

The outcome. The rep walks into every first meeting already knowing who’s on the call and why they might care, so the first ten minutes are about the buyer’s problem rather than discovering it.

The sources. All of these exist for almost every first meeting, and almost nobody reads them, because reading them takes about ninety minutes per meeting:

  1. Your own CRM history with the account: past conversations, past opportunities, who touched it and what happened.
  2. The attendees’ public professional profiles: role, how long they’ve been in it, where they came from.
  3. What they’ve been posting or speaking about recently.
  4. Company news from the last quarter: funding, leadership changes, incidents, acquisitions, layoffs.
  5. What changed in their technology stack, where it’s visible.
  6. Whether they’ve publicly complained about their current vendor.
  7. Your competitive library entry for whoever they use today.

The output. One page, delivered the evening before. Not a dossier. Five short sections: who’s in the room, what they probably care about this quarter, the one thing they’re likely to push back on, the question you should ask in the first five minutes, and what not to say.

The format rule. If it’s longer than a page, the rep won’t read it, and a brief nobody reads is a crutch with extra steps.

The setup. An afternoon to write the instructions, then ongoing tuning as reps tell you what was useful and what was noise. The per-meeting cost is small enough not to matter.

The failure modes. Three, and you should design for all of them.

  1. Confidently wrong. A model will happily connect a person to the wrong employer or a stale title. Every claim in the brief needs a source, and the rep needs to know which parts to verify.
  2. Creepy. There’s a line between prepared and invasive, and a buyer can feel it. The brief is for the rep’s understanding, not for reciting. “I saw you posted about this” is fine. Reading somebody’s last six months back to them is not.
  3. Generic. If the brief says the same thing about every account, it’s the category mean again. The test is whether two briefs for two different companies would be obviously different.

Who owns this? Marketing, now. Because the outcome is the meeting, and marketing is the function with the research skills, the competitive library, and the content to put in it. That’s the redesign. Not a new tool. A new line around what counts as done.


What should never be automated

The flip side, because a chapter that only says “automate more” is half a chapter.

  1. The decision about what’s true. A model can assemble the evidence. A person decides what the company claims.
  2. The decision about what to say, and who says it. Point of view is the scarce thing in this category, and it can’t be generated. It can only be held.
  3. The relationship. Nobody has ever been loyal to a sequence.
  4. The apology. If something went wrong, a human says so, in their own words.
  5. The judgment at the end. Humans on the loop, not out of it. Every system in Chapter 15 ends with a person choosing.

And one more failure mode, which is the most common one I see: automating a process nobody should be running. If the gated report sent to a cold list didn’t work by hand, it won’t work automated, it’ll just fail at a scale where the failure looks like activity. Run the Comfort Check before you automate anything. A process that fails question two shouldn’t be made faster. It should be stopped.


The Comfort Check

Chapter 17: the handoff

1. The comfortable assumption. Marketing’s job ends when the lead reaches sales.

2. The true objective. A first meeting worth the buyer’s time.

3. The redesigned approach. Marketing owns the brief behind every first meeting: seven sources, one page, delivered the night before, with a human checking what’s true.