The 30% Jerk

Why unpredictable people — and unpredictable tools — create unpredictable stress and cost you more than difficult ones ever will

I was on the phone with my coach a few days ago, circling a situation that had been living rent-free in my head for longer than I care to admit, and I heard myself say something that stopped us both.

“The problem isn’t that he’s a jerk. The problem is that he’s only a jerk about 30%  of the time, but when he is, look out!”

For the record, “jerk” was not the word I used….but that’s not the point here.

Let’s move on.

The more I sat with the realization of how painful and draining this relationship was for me, the more I realized I had stumbled into one of the most underappreciated truths in stress science… and one that explains a lot more than one complex relationship.

A consistent jerk is not a stressor. A consistent jerk is a predictable pattern.

Think about the person in your life who is reliably prickly. Every meeting, every email, every text and every in-person exchange, they bring the same energy. Is it pleasant? No. But notice what your nervous system does with it: almost nothing. You keep the exchange short. You have a model, the model works, and you spend exactly zero calories monitoring the situation in between.

Now think about the person who is warm and generous 70%  of the time… and then, without warning, isn’t.

That one you think about more than you probably care to. It’s the relationship that haunts you as you try to fall asleep at night.

You rehearse the conversation before you walk in. You reread your own email looking for the thing that might set it off. You keep a low, humming background process running that never fully shuts down, because it can’t. There’s no consistent pattern that would let you know what’s coming. That’s not a personality problem. That’s a prediction problem, and your brain, which is fundamentally a prediction machine, absolutely HATES it.

In 2016, researchers at University College London published a study in Nature Communications that I think about constantly. They put 45 volunteers in front of a simple computer game: turn over rocks, some of which have snakes underneath, and a snake means a mild electric shock to the hand. As people played, the researchers tracked their stress. 

Here’s what they found. Stress did not peak when the chance of a shock was 100%. It peaked when the chance was 50%.

As lead author Archy de Berker put it: “It turns out that it’s much worse not knowing you are going to get a shock than knowing you definitely will or won’t.”

It’s the coin flip that wrecks you.

And this tracks a much older finding. Jay Weiss demonstrated back in 1970 that rats receiving unpredictable shocks developed dramatically more stomach ulceration than rats receiving the exact same shocks preceded by a warning tone. Same voltage. Same duration. Wildly different bodies at the end of it.

The dose of stress wasn’t what did the damage.

The unpredictability of the dose did the damage.

Which brings me to the machine on your desk

I read a survey last week that read completely differently once I had the 30% jerk in my head.

BambooHR surveyed 1,608 full-time U.S. workers this summer about how they actually spend their time with AI tools. The average worker reported 87 minutes a day inside them — roughly 47 workdays a year. But only about 35% of that time went to productive work. Some 42% went to troubleshooting errors and re-writing prompts, which the report translates into about 20 workdays a year lost to arguing with the thing that was supposed to save you time.

Every headline I saw ran the same take: AI productivity is a myth, the emperor has no clothes, and so on.

I think that’s the small story.

Here’s the big one. Your AI tools are a 30% jerk.

They are brilliant most of the time. And then, unpredictably, confidently, and with no warning tone whatsoever, they are wrong, and you cannot build a reliable model of when that will happen. That means you can never actually put down the vigilance.

That is not just lost output.

That is continuous low-grade threat monitoring, which is a metabolically different animal….. and it requires that you recover differently too.

Hard work makes you tired.

Unresolvable uncertainty makes you depleted, and those are not the same.

Everything is unpredictable and unpredictability is the one thing your dashboard was never built to measure.

The cheapest resilience intervention nobody budgets for

We spend an extraordinary amount of money trying to lower the stress load in our organizations.

Reduce the demands.

Add the wellness app.

Shorten the meeting.

But the research keeps pointing somewhere less expensive and much less glamorous. It’s not the size of the stress load that breaks people. It’s the randomness. Stress you can see coming can be converted and used as fuel. Stress you can’t predict feels more like erosion.

You cannot make the 30% jerk predictable. I’ve tried. You’ve tried. That’s the whole reason the vigilance never shuts off. But go back to Weiss’s rats for a second, because the solution is right there in front of us. The warned rats did not get fewer shocks. Same voltage, same duration, same total misery delivered. What they got was a tone.

And notice what that tone actually was. It wasn’t information about content.  The rats never learned what was coming, or how bad it would be.

The tone only told them WHEN to expect the shock.

That’s the mechanism. Predictability is mostly a question of timing. The tone’s real gift was never the heads-up before….it was that the silence became information. Silence meant off. And a rat who knows when it’s off duty gets to actually be off duty.

They weren’t braced better. They were off-duty more.

That’s where your damage is happening.

Not in the 30%.

In the 70%.

And I want to be precise about that, because the 70% is the part that looks fine from the outside. It’s the warm stretch. The good weeks. The meetings that went well. It is also every hour you spent braced for the other thing — which is how you can have a perfectly pleasant month and end it depleted.

So let’s sit with that for a moment in your own life.

How many hours a day does the fear of AI occupy you? Not how long you worked on it or with it. How long it has you thinking, worrying, or pushing back?

(Anyone else noticing that the gap is bigger than the work?)

I have watched people start stressing on a Sunday night about being behind on AI when the tool doesn’t get introduced until the following Tuesday. I have BEEN that person. Days of physiology spent on a threat that isn’t in the room yet, hasn’t been scheduled, and likely won’t even show up in the form you’re rehearsing for.

Nobody handed you a tone…but luckily, you aren’t the rat. You’re the human in this experiment, and you can schedule a tone yourself.

Pick the window. “I deal with AI stuff between ten and eleven, Monday through Friday.” Write it somewhere you’ll see it. It does not have to be generous. It has to be REAL, and it has to have edges.

Then every hour outside that window is certified safe. Off. The way Saturday is certified for you now, if you’re lucky enough to have a job that still has a Saturday in it.

I’m not suggesting you can flip a switch and turn the worry off. I’m suggesting we acknowledge it like a passing cloud. Catch yourself in it.  Let the fear be acknowledged, and then let it go like a cloud passing through. Protect the time when you get to be OFF. It will feel fake at first.  Your prediction machine doesn’t trust a new rule until it’s run the experiment a few times and found the building still standing.

Which is the same thing I say about everything else. The story doesn’t update until you take the action. Give it three weeks of real evidence before you decide it didn’t work.

And here’s why that worry felt so necessary you never thought to question it. It feels like preparation. It has the exact texture of work. But sitting up Sunday night running scenarios about Tuesday produces precisely zero information you didn’t already have. Tuesday produces information. Sunday produces cortisol. Not the same thing.

This is the cheapest thing in this entire post. No vendor. No line item. No pilot program, no committee, no app. It’s a boundary with a time on it.

You are allowed to be off duty. Not when the uncertainty resolves, because it isn’t going to — not this year, probably not next.

And there is something almost freeing in that, once you sit with it. We have been doing this our whole lives. Every one of us has already lived through a dozen versions of the thing that was definitely going to be different this time.

You do not need the uncertainty to end before you’re allowed to put it down.

You’re allowed to be off duty on a Wednesday at six because you decided that Wednesday at six is off. 

Stay stressed, friends. Just learn to fear it a bit less.

Xoxo Rebecca