Put "Email" Where "AI" Was

The Financial Times ran a piece recently (Registration Wall) about companies that pushed their staff to use AI, and how that has backfired. Amazon set usage targets for most of its developers and posted the numbers on internal leaderboards. Then it took the leaderboards down, after the FT reported that people were automating junk work to climb the rankings. An Amazon senior vice president told staff: "Please don't use AI just for the sake of using AI."

I replaced every "AI" in that article with "email."

I do this a lot. AI and email are not the same thing, and I'll get to where that comparison falls apart. It's still the fastest way I can triage whether someone has their head wrapped around AI and is talking about it the "right" way (yes, I know how that sounds, but there I said it). Most of the time, if someone is talking about AI - which is a tool - a different tool should fit in that sentence. Email, fax, a web browser. It gets funnier, sure, but it should still make sense. When it stops making sense completely, something is off.

The Substitutions

A chief executive worried his engineers were using too few Email tokens. "If an engineer being paid $500,000 a year did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed."

Before the mandates came, workers "worried first about being caught cheating by using the powerful new tools. They were more likely to conceal their usage than boast about it." Two years ago people hid their email in shame. Now they're ranked on it.

A law firm offered a £1mn bonus if its 1,200 staff could send 1mn emails in a year. They hit it with four months to spare. Nobody seems to have asked what was in the emails.

A consultancy said workers who couldn't upgrade their skills for the Email age would be "exited."

A professional services firm built a dashboard where staff compare their email usage, and runs a quarterly "Email Spark" award for the brightest Email innovations.

The push to hand more tasks to Email has unsettled workers, "and not only because they fear it threatens their own jobs."

And one engineer, unhappy about Email usage showing up in his review: "It removes my choice . . . I don't feel I should be penalised for not using a tool I don't feel like using."

Why I Start From a Tool

Start from "this is unprecedented" and you've handed out the exemption before anyone earned it. Watch what rides in behind it: we can't show a return yet, normal measurement doesn't apply, we'll require it now and sort out the value later, we'll give a prize to whoever uses it most. All of that follows from granting specialness up front.

Start from a tool and the questions get boring again. What did it cost? What came back? Who's better off? Same questions you'd ask about a forklift.

Swapping the word is how I check whether someone's characterization of AI makes sense, and how much of it does. Email is my go-to. Put an old apple where the new one is and see if the sentence survives.

Here's what it comes down to. If the sentence still makes sense with email in it, the person has AI in the right context. Whatever they've hung on it - a truth, a fear, a prediction, a use case, a product claim - gets a full hearing from me. No discount.

If it sounds silly, something is off, and there are only two ways it can be off. Either they're pointing at something AI actually does that email can't, which is a fair break and usually the start of a better conversation. Or somebody is selling something.

Where the Swap Is Fair

Sometimes it catches something that doesn't make sense. Once the word AI is gone, what's off is easy to name.

Counting use of the tool instead of the work - "Staff logged 1 million prompts" becomes "staff sent 1 million emails." A million emails isn't an achievement. Nobody won a customer or shipped anything faster because a number went up.

Calling policy and training a strategy - "We need an AI strategy" becomes "we need an email strategy." You did need something for email. Rules about what not to put in writing, training so nobody replied-all to 12,000 people, a retention policy, a lawyer's opinion. That's governance. Point it at customers, like handling service requests over email, and now you do need a strategy - same with the AI version. Pointed at your own staff at their desks, it's training with a bigger word on it.

When someone tries to avoid accountability - "The AI made it up" becomes "the email must have gone to spam." Both blame the pipe. Somebody set that thing up and somebody owns what comes out of it.

Advice from whoever sells it - "Your competitors are already using AI" becomes "your competitors are already using email." The person telling you to use more of it is billing you for the usage.

Fear with nothing behind it - "AI is coming for us" becomes "email is coming for us." If a person can't say what it does, to which job, they're scared of a word.

The word as a badge - "We're an AI-first company" becomes "we're an email-first company." That one's aimed at investors and job seekers.

Where the Swap Stops Being Fair

Use it to wave off something the tool actually does and it turns into a dodge immediately. If someone says this drafts our first-pass contracts, "relax, it's just a tool, like email" isn't an answer. The swap can tell you when someone is talking nonsense. It can't measure what AI actually does to the thing you're talking about.

Someone who has never used AI can hide behind this line forever. Working out what it does to a person's job takes effort, and saying what you found takes some nerve, because it might not be what anyone wants to hear. The line skips all of that. It's worst pointed at the person whose job is actually on the line - telling a paralegal that people panicked about email too is easy when it isn't your job, and the typing pools never came back.

And it falls apart if you push it to "so nothing will happen." The swap only works because email sounds boring today. Email was never boring. It changed who talks to whom, wiped out whole job categories, and followed people home at night. Anything that proves AI won't amount to much proves email didn't either, and email rebuilt how companies work.

The Ones That Hold Up

Plenty of people in that article came through the swap without a scratch, because they were already treating it as a tool.

• Before:* "Reward staff for being effective and efficient and having more output, not for necessarily using AI."
• After:* "Reward staff for being effective and efficient and having more output, not for necessarily using email."

Nothing changes. That's a statement about how you pay people, and it was true before anyone had AI.

• Before:* "Our estimate would be that this AI tool should save you time. Could you tell us why you aren't using it?"
• After:* "Our estimate would be that this email tool should save you time. Could you tell us why you aren't using it?"

Reasonable either way. It's a question rather than an order, and it asks about time saved instead of usage.

• Before:* "Now we need to see a return on it, rather than spraying and praying."
• After:* Identical. The word never appears. It's asking what came back for what went out.

What AI Does That Email Can't

Fax was a faster letter. Email was a faster memo. A browser was a faster library. Every office tool I can think of showed up as a quicker version of something we already did, which is why swapping one for another works as well as it does.

This one has no obvious ancestor. The closest thing to it is a person. That's why the test only gets you about seventy percent of the way, and why the last thirty is where all the argument actually is: the machine says things that can be wrong, it costs money every time you use it, and it does some of the work instead of carrying it.

So run the swap. See what falls over. What's left standing is the conversation that needs having - the risks, the benefits, the things nobody has settled yet. What fell was noise, hype, fear, and somebody selling something. That pile isn't harmless. A lot of the anxiety people carry about AI came out of it.