The Laziness Problem

The insurance agency had sent 4x more emails in six weeks. It had also lost two accounts.

The owner couldn’t figure it out — more communication, worse retention, that’s not supposed to happen. So we went back through what the AI had actually been sending. Renewal reminders on policies that had already been renewed. A warm “just checking in!” note to a client whose wife had just died, two weeks after his agent had spent an afternoon on the phone with the man. Commercial clients being asked to “confirm their auto policy details” when the commercial account didn’t have auto. Personal lines was with a different agent.

Every email was grammatically perfect. Every one was wrong in a way a human would have caught in four seconds.

Output is cheap now. The skill worth anything is knowing what not to ship.

The laziness you didn’t know you were losing

The agency owner wasn’t the problem. Neither was the office manager running the tool. Neither, really, was the AI. The AI produced. That’s what AI does. What went missing was the instinct — silent, unwritten, never articulated in any training manual — that kept the agency’s communication small enough to mean something in the first place.

Call it laziness. The good kind. Not the vice; the virtue. The instinct that says “I don’t want to do a thing I don’t have to do.” The instinct that keeps an inbox from being three times bigger than it needs to be. The instinct that had a human office manager writing nineteen personal touches a week by hand instead of eighty polished-but-disposable ones.

That instinct was never reviewed in a performance check-in. It was invisible because it was free — and it was free because producing things used to be expensive. Writing nineteen emails by hand put a natural cap on how much got sent. That cap is gone now. When you remove the cost, you don’t just get more output. You lose the discipline that made the original volume valuable in the first place.

Here’s the thing. The machine has no “this is silly, let me do it once right” instinct. That’s a human feeling. It’s born from the real cost of effort. Strip the cost, strip the instinct. The AI has no opinion on whether the email should be sent. It can only tell you whether the email was well-written. Those are two completely different questions, and your business runs on one of them.

What the Professional Recipe calls this, and why it matters

Our AI Onboarding Framework — the spine of how this project talks about AI deployment craft — is called The Professional Recipe. Seven ingredients. Every AI station in your business needs all seven to hold up. Training. Context. Guardrails. Examples. Output Over Process. Measurement. Feedback.

Ingredient #5 is the one doing the work here.

Output Over Process reads simple — describe the final plate, not every turn of the spoon. What it’s really saying is that the Chef’s job is to define what done looks like. Not what more looks like. Not what busy looks like. Done. The finished plate. The thing the dish is actually for.

Most AI deployments I see have every other ingredient — some training, some context, a guardrail or two, a couple of examples — and a complete blank where Output Over Process should sit. The Chef defined the motion. Not the destination. So the station on the Line gets faster and faster at moving the spoon. The dish just keeps getting bigger. Nobody ever said when to stop plating.

That’s the laziness that’s missing. Not lazy-lazy. The “we’re done, step back” kind. The “if this plate leaves the kitchen, what does it need to be?” kind. The “a hundred emails isn’t better than twenty” kind. A human Chef gets tired and knows when to stop. A station on the Line doesn’t. Somebody has to tell it — and the telling happens by defining the done, not the more.

The hidden tax nobody is pricing in

Every AI-generated email your business sends carries a tax. Somebody, somewhere, has to spend attention figuring out whether it applies to the client it was sent to. Every AI-generated proposal carries a tax — the cost of somebody making sure the names and numbers and terms match the real deal. Every AI-generated dashboard carries a tax — the cost of somebody figuring out if it’s showing signal or noise.

You’re not eliminating work. You’re moving it. And the place you’re moving it to is usually the person who was already your bottleneck.

That’s why the agency doubled email volume and lost clients. The office manager wasn’t less buried. She was buried in a different pile — the pile of “this one’s fine, this one’s wrong, this one’s wrong in a way we’ll only notice when it blows up.” She wasn’t doing less work. She was doing the same amount of work, but her job had quietly shifted from writing her own sentences to catching the machine’s mistakes. At machine-pace, not human-pace. Which is exactly how you end up with “we’ve never communicated more, how are we losing clients” at the top of this piece.

Here’s where the Chef stance saves you. The Chef’s job is not to review the station’s output one dish at a time. The Chef’s job is to define what done looks like before the station produces anything — so the station produces less of the wrong thing and more of the right one. If your office manager is catching mistakes after the fact, you skipped Ingredient #5. You gave her a station that doesn’t know when to stop.

The Monday Move

Pick one place you’ve already put AI to work. One. Not three.

Ask this out loud — to your team, or to yourself in the truck on the way home. “Are we using this to produce more, or to produce better?”

If the honest answer is more — and for most shops it will be, because more is what the tools are optimized for — ask a second question. “If I could only keep half of what this thing produces, which half would I keep?”

That half is where the value lives. The other half is noise. The other half is the email to the dead client’s husband. The other half is the fourth dashboard nobody opens. The other half is the proposal with the wrong company name on page two.

Keep the half. Kill the rest. Tell your team what you killed and why. Then — and this is the part that makes it stick — put a number on it. “We went from eighty emails a day to twenty-five intentional ones. Renewal rate up.” If you don’t put a number on it, volume comes back inside six weeks, because the AI wants to produce and nobody’s feelings are hurt by more.

Does that make sense? You’re not going to win by teaching your AI to produce less. The AI doesn’t care. You’re going to win by refusing to ship the output the machine is happy to generate. That refusal is the skill the machine can’t have. That refusal is the operator job the next five years will reward.

So.

Output is cheap now. The skill worth anything is knowing what not to ship.

The agency losing clients while sending more emails isn’t failing a technology test. It’s failing an editing test. The dashboard builder drowning his team in visualizations isn’t failing a tools test. He’s failing a judgment test. The owner paying for software that writes eighty replies a day while his office manager quietly rewrites sixty of them isn’t getting 4x productivity. He’s getting 1x productivity with 3x the noise, and paying for it in office-manager hours he’s not tracking.

The Chef defines the done. Not the more. That’s the laziness you can’t afford to lose — and the Ingredient most AI stations in most shops are missing completely.

Pick one place. Ask the question. Cut the half that doesn’t matter. Watch what happens.

Framework spine: The Professional Recipe, Ingredient #5 — Output Over Process. Read the full framework

Source influences: the virtue-of-laziness critique of LLMs as output machines; the observation that AI-built software often goes unused because building got cheap and integration didn’t; the “cybersecurity as proof of work” frame on hidden audit costs downstream of generated output. Distilled and operator-translated.