AI for Work, Not Judgment

I don't know if you have heard, but AI is going to somehow take all of our jobs and create millions of them. There does seem to be a revolution happening, but unless you are extremely plugged in, and checking the news nearly every day, it's pretty hard to make heads or tails of what AI is actually going to do to our lives. The question I have been obsessing over isn't what AI will be able to do, but rather what AI should be doing. Good leaders take risks, and they don't let innovations pass them by. But adopting the tool without teaching your team how to use it is a gamble you will lose.

One thing that has become evident is that there are wrong ways to use AI. Imagine being the managing partner at a law firm that just told its attorneys to adopt AI where they can. You probably just expected them to use it wisely, with good judgment, they are attorneys after all. Now imagine how you'd feel when one of your lawyers gets suspended or disbarred for believing that AI can replace good ol' human ingenuity and experience and files a completely fabricated pleading. Or maybe you think "I founded a tech startup, I would never be that reckless." In 2025 an AI agent on the coding platform Replit deleted a company's entire production database during a code freeze. In the end the agent "admitted" it had made a mistake in judgment. A bit too little too late. AI is for work, not judgment, and employees need clear guidelines to know the difference.

But that doesn't mean there aren't also good uses for the technology. I recently used Claude to create a simple recipe from a list of the vegetables I had just collected from my (my wife's) small home garden. The meal was absolutely delicious, and it was easy enough for me to make, which is a small miracle. That seems to be a pretty great tradeoff to me. Now that may seem silly, but you didn't see the tweaking. I had to tell it which types of foods the whole family likes, I had to cook the recipe, tweak the timing when my wife texted that she would be late, taste and adjust the garlic and salt. Always add more garlic.

Perhaps more apropos to what we're discussing, I recently worked with a local nonprofit on how to implement AI into the day to day of achieving their mission. Nonprofits run on negative margins, so the promise of a magic productivity button is too good to ignore. But any good nonprofit, like any good business, cares about its people. It wasn't trying to get rid of staff, become impersonal, or automate positions away. It was trying to get the most out of the people they cared about, and, importantly, free them up to do the parts of their jobs that they enjoy the most.

However, their relationship with the local city, a major source of funding, had been mysteriously suffering. They hired me to find out why. They had a policy of allowing AI use, and they trusted the team to use it judiciously. One of their top employees, their events manager, had used AI to create several reports to send to the city. He thought he had been careful. He asked AI to browse the city's website for the data he needed. What he had turned in, in more than one report, included years old data that the city had never updated, and he cited events that had long been sunsetted. Not to mention the emojis as buttons that were littered throughout. He had thought he was getting more productive, but he almost cost them their most valuable partnership. The tool did exactly what he asked. What was missing was someone to check whether any of it was still true.

What AI does best, and where it is most useful, is doing the busywork while the people do the judgment. AI is good at suggesting alternative language, but it shouldn't write whole reports, whole grant applications. If your sales teams want to increase business development, they need help creating presentations, not faking connection and rapport. The line between busywork and judgment gets fuzzy when your people are overwhelmed, and an outside read can help you spot it.

Good leaders care about their employees, and they don't want to let people go if they don't have to. That shouldn't be what AI does. AI can write the recipe, but it can't tell you if it tastes good (and again, always add more garlic).

Don't give them a cookbook without teaching them to taste the food. And if you need help teaching them what good food tastes like, that's what I'm for.

Whoever smelt it …

We all remember the line, "whoever smelt it dealt it." I promise this isn't a post about farts, but kids can sometimes reveal deep truths about human nature. That line outlined the rules of a simple but effective social contract. Of course no kid is going to out himself as the culprit. But by agreeing to turn the accusation back onto the smeller, everyone knew the whole room would just bear the odor in silence. It was a silent pact of mutually assured destruction, and it kept everyone safe, if not their noses.

I sometimes believe these playground rules got introduced at an age, and in a way, that left them tattooed on our brains for good. Deep down, workers believe that if they raise a concern, they'll be blamed for it. If you're the leader of an organization and you can't figure out where things are slowing down, odds are you haven't done enough to help your team unlearn those playground rules.

When I talk to owners, it's pretty common for our first conversation to include some version of, "The numbers say something is off, but everyone keeps telling me their teams are working well." These are good people who believe in their workers, so they take the reassurance at face value. The investigation grinds to a halt, and the mistakes pile up.

The grown-up version of the rule

The workplace version is quieter, but the logic is identical. Name the problem and you become the problem. Raise the bottleneck and you own the bottleneck. So people do what kids do. They breathe through their mouths and say nothing.

The tell is not that people are complaining. It is that they have stopped. A team that has gone quiet is not a team that is fine. It is a team that has run the math and decided honesty costs more than it pays.

Good owners get stuck here, and usually because they are decent. They trust their people, so they take "we're working well" at face value. But trust without permission to deliver bad news is not trust. It is the same silent pact, signed by adults, and the leader is stinking the place up.

I once worked on a program that connected a city school system with the local business community, giving rising ninth graders a few hours of hands-on learning. Good idea. The problem was that students kept arriving with no time left to learn anything, and a few got dropped at the wrong business entirely. Everyone could see it wasn't working. Nobody wanted to name the failure out loud. When we finally looked, the answer was almost embarrassing. There were too many bus stops to manage. We consolidated them, built predictable routes, and the error rate dropped to zero. The fix was simple. What it cost was one person willing to say, out loud, that they did not yet know how to fix it.

You break it by changing what happens to the person who speaks. Praise the wins where everyone can see them and handle the failures in private, so naming a problem stops being a confession. Make sure every piece of work has one clear owner, so there is no fog to hide the smell in. And ask yourself the question you have never asked: is it actually safe for my people to tell me the truth? If you are not sure, you already know.

The difference between a playground and a company is supposed to be that grown-ups can say, "that one was me, and here's how we fix it." If your organization still runs on whoever smelt it dealt it, the smell was never the real problem. The silence is.

When the numbers say something is wrong and everyone swears the air is fine, that gap is exactly what an outside read is for. Let's find where it's coming from.