Chapter 14 · Part Four: The Machine That Makes Average
There are two ways to use AI at work, and on a status report they look identical.
A crutch does the work you were already going to do, faster.
A multiplier does work you could not have done at all.
Both show up in the quarterly update as “we adopted AI.” Both save time. Both make a nice slide. One of them makes you better at your job. The other makes you faster at being interchangeable.
Why you can’t tell them apart from the inside
The crutch feels great. That’s the problem.
Writing the blog post in forty minutes instead of four hours is a real, immediate, personal relief. You feel it the same afternoon. The multiplier, by contrast, usually feels like more work at first, because it’s doing something that didn’t exist before and somebody has to figure out what to do with the output.
So left alone, every team drifts toward crutches. Not because anyone is lazy, but because the reward for a crutch arrives today and the reward for a multiplier arrives next quarter. That’s the same asymmetry as Chapter 3, wearing a different jacket.
And the metrics most companies use to track AI adoption make it worse. Seats licensed. Prompts per week. Hours saved. Every one of those numbers goes up just as fast for a crutch as for a multiplier, and some of them go up faster.
The six-month test
There’s exactly one test I trust, and it isn’t adoption, hours saved, or seats.
Six months in, is your work more distinctive, or more average?
Run it on your own output. Take what your team shipped six months ago and what it shipped last month. Put them next to the competition, not next to each other. Has the gap between you and the category gotten wider or narrower?
If it’s narrower, you bought a crutch, and the time it saved was spent producing more of the middle. If it’s wider, you have a multiplier, and you should find out exactly which part of the workflow is responsible and protect it.
The test has a second version for people. Six months in, is this person doing work they could not have done before, or the same work faster? The first is a career. The second is a countdown.
A taxonomy
Here’s how I’d call the common uses in a marketing function. These are judgments, not laws, and a few of them flip depending on how you run them.
| Use | Call | Why |
|---|---|---|
| Drafting blog posts from a topic | Crutch | It produces the category’s post on the topic, which is the post your competitors are also producing |
| Drafting from your own primary material (calls, interviews, internal debates) | Multiplier | The input is unique, so the output can be |
| Generating social posts at volume | Crutch | More of the same, faster, in a feed that already has plenty |
| Finding the stories worth a point of view, then letting a human decide | Multiplier | It reads more than any person can and leaves the judgment where it belongs |
| Summarizing meetings and calls | Conditional | Crutch if the summary goes in a folder. Multiplier if it’s mined for what customers actually say, across hundreds of calls |
| Rewriting copy to “make it better” | Crutch | Better, to a model, means smoother. Smoother means closer to the mean |
| Adversarial critique of a draft | Multiplier | It adds friction, which is the scarce thing |
| Researching an account before a first meeting | Multiplier | Nobody was doing it properly before, because it took ninety minutes per meeting |
| Writing in an executive’s voice from a real profile | Conditional | Multiplier if the ideas are theirs and the time goes back to them. Crutch if it becomes a way to post without having anything to say |
| Competitive monitoring on a schedule | Multiplier | The alternative wasn’t a slower version. It was a spreadsheet nobody updated |
| Image generation for campaign creative | Conditional | Crutch for stock-style images, which is the hooded figure again. Multiplier for testing ten strange concepts before you commission one |
| Reporting and dashboards | Conditional | Crutch if it makes the same report faster. Multiplier if it produces a read nobody asked for that changes a decision |
| Personalized outbound at scale | Crutch, usually | Personalization that a buyer can see was generated is worse than none |
The pattern in the multiplier column is consistent. A multiplier either works from input nobody else has, does something nobody was doing at all, or adds friction where the process had none. A crutch takes the thing you were already producing and produces it with less effort.
Making it on purpose
The useful question isn’t “how do we adopt AI?” It’s “what could we do now that we couldn’t do before?” Ask it about the specific work your team does, and write down the answers before you buy anything.
Most of the answers will be things that took too long for anybody to do well. Research before every meeting. Reading every call transcript. Tracking every competitor every week. Answering the same question from every new rep. The crutch version of AI makes your existing work cheaper. The multiplier version makes impossible work possible, and that’s the only version that changes your position.
The next chapter is what that looks like when you actually build it.
The Comfort Check
Chapter 14: AI adoption
1. The comfortable assumption. We’re doing well with AI, because adoption and hours saved are up.
2. The true objective. Do work that was impossible before, and widen the gap between us and the category.
3. The redesigned approach. Judge every use by the six-month test, keep the ones that work from unique input or add friction, and stop counting seats.