Why AI Makes Good Teams Better and Bad Teams Worse

It was 11pm on a Tuesday in winter. The kids were asleep, the house was quiet, and I was on the couch in front of the fireplace, building.

That night it was a lead generation system. I watched Claude work inside Chrome, using Sales Navigator the way a person does: running searches, building lists, adding people, personalizing the notes. Doing the work, like an SDR sitting next to me.

I thought, uh-oh. This is incredible. And I kept going for hours.

That was the night I noticed the dark side of AI.

Over the months that followed, I worked out the one thing every leader needs to understand before rolling this out. AI amplified something that was already in me. The real work is deciding what you want it to amplify before you hand the tools to your team.

The hyper-productivity trap

Some days, some weeks, even some months, you fall into hyper-productivity. The admin is done in minutes. So you keep going, sometimes until 1am or 2am. You think, let me build one more feature. And because there is a little shot of adrenaline every time the terminal says "done," you want to see the next one.

I have caught myself in that state for weeks at a time. Especially in winter, in a beach town, when there is not much going on. You get a few hits of that feeling and you think nobody can stop me. You feel superhuman.

So you trade sleep and skip the workout. The next morning you are not as sharp, and you miss the quiet time with the kids. Your mood gets anxious, always wanting to build, a little less social with humans and a little more social with the machine. My laptop goes everywhere now, always cracked open because there is a session running.

Let’s face it: Claude Code in the terminal is addictive, and I think it is most addictive for exactly the people who are usually the best to hire.

Curious people with a growth mindset, the ones who want to achieve more, and then want more again.

For the first time, if you are not a developer, you can just talk to the machine and watch it build. It is not lonely like writing code alone. There is an intelligence working beside you, any hour of the day or night.

That is the gift. It is also the trap.

You are always building, because you can. The real question is whether you should.

AI is an amplifier

Here is the thing I had to learn the hard way. AI is an amplifier. The technology is neutral, and your culture decides what it amplifies.

Give it to a proactive team. People with curiosity and grit, who do not wait for Monday's meeting to know what to do. AI multiplies all of it, massively.

Give the same AI to a reactive team that does the minimum, and it changes very little. They ask it to do the task, it gets done, and they sit back down.

You can hear the difference in how people talk to it. A fixed mindset says, "I need this, can you just do it," and accepts the first thing that comes back. A growth mindset says, "We need this, how should we do it, what do you need from me?" That second conversation is slower, and it almost always produces something better.

I saw this on a call recently with a team inside a larger company. A small reporting task gave real value to clients, but took so long by hand they had quietly stopped doing it for most accounts. Doing it for everyone would have meant four or five hires they could not afford. During the program, they realized AI could run it. Now one person does it for every client.

The culture AI punishes is the hidden, low-trust one, where people do the minimum and cover it up. Drop powerful AI into that and it gets worse, faster.

So AI will amplify your culture, good or bad. That is the most important thing a leader needs to understand before rolling any of this out.

More is not better

Here is the cleanest version of this lesson: a mistake I made with a client.

Several people were involved in the decision, and I wanted to impress all of them. So I built some extra reports for one of the other leaders. AI helped me map it out, I got excited, and I built a full application for it.

Then the person told me, plainly, "I never asked you for that." They already had better data than I did, straight from their own systems while mine came from surveys. The work was good. It was just unwanted.

It cost me a week of back and forth and an application that went nowhere. Because you can build it does not mean you should. Sometimes people want to take their time, for reasons that make sense in their world.

You also see two kinds of people in that moment. One looks at a rough prototype and asks about security, authentication, and everything that is missing. The other says, that is amazing, you built it in no time, now how do we make it safe.

The first will slow everything down. The second will fly.

This is exactly why ADOPT starts with Align. Most of us jump straight to building, but alignment is where it should start.

The framework: Work, Self, Team

The key is to be deliberate. It comes down to three things: the work you hand off, the boundaries you set, and the people you hire.

Work: let the agents handle the grind

Hand the machine two kinds of grind.

The first is the boring, repetitive work. My sales agent lives in Sales Navigator. It pulls people from saved searches into the right list, moves them once they connect, tracks them, and writes the details into the CRM.

The second is the work you would never get to, because it would eat your whole night. The same agent reads through hundreds of discovery call transcripts and pulls out patterns I would never find by hand. My SEO agent checks Search Console every day, works the keyword data, and watches what competitors rank for. That used to be a someday task. Now it just happens.

My rule for what an agent can own is simple. Anything that goes out in public, an email, a LinkedIn DM, a post, the newsletter, a blog, even a message to a colleague, I review before it sends. Anything internal, research, organizing notes in Notion, tagging, pulling insight out of a transcript, runs on its own.

Now the honest part. Handing off the grind gave me time back, and I refilled it with more clients, more building, more launching. Where you carried two or three clients, you can carry ten. I will not pretend I bought that time back to rest. The point is that the time comes back, and you get to choose what it is for.

Self: set the boundaries before you need them

I am an always-on person. So this section is not "log off at six." That would be a lie. It is the few rules that keep hyper-productivity from tipping back into the 2am, laptop-always-open, missed-the-morning version.

Eight hours of sleep is non-negotiable. Dinner with the kids most nights, and when a US call gets in the way, I am there for story time and bed instead. Two late nights a week are allowed, for building or for fun. The rest, I want to be in bed before midnight.

My Oura ring keeps me honest about whether the late nights are worth it.

The hardest rule to hold is closing the laptop by 10 and keeping the last hour screen-free. I know that is the healthy thing. I break it constantly right now. And I have made peace with this: if the trend is good more often than not, that is okay. Do not be cruel to yourself.

The line I am protecting is simple: sleep, my body, presence with my kids, the gym, a clear head.

Team: build the culture before the tools

This one is mostly about who you hire and what you are all building toward. Tools come second.

AI amplifies the team you already have. A team that trusts each other and shares a mission gets faster and bolder. A team built on pressure and low trust just runs its dysfunction more efficiently.

I have seen this on the client side. Out of many cohorts, one stands out as the one that went wrong. The leadership wanted pure output. Ten times the work, more revenue, full stop.

Our early surveys showed the team was grateful and changing how they thought about their work, and we could show a real return. The leadership cancelled anyway, and not long after, a chunk of those same people lost their jobs.

Leaders who expect 10x from one person while letting nine others go are not building the kind of culture this is for.

So when I coach leaders rolling out AI, the warning is the same. If the only goal is more output, AI will give you more output and a team that quietly burns out and leaves. Culture before tools. Every time.

If you want an honest read on what AI would amplify in your team before you roll it out, that is what our AI Audit is for.

Who is Responsible for AI's Output?

The concerns people raise in our programs are real. Weaker judgment. Thinking less for yourself. If you get lazy and let AI run without checking it twice, you will ship mistakes and look unprofessional.

The bottom line: AI is not responsible for its output. You are.

This is why I keep pushing Human plus AI instead of AI on its own. A lot of roles only exist because machines could not do the task. People copied and pasted between spreadsheets and emails. They built the same document by hand a hundred times. That work is going away, and honestly, it should.

By the end of 2026, one person will reasonably do the work of ten, and the models keep getting stronger: smarter, and able to work for longer stretches on their own.

The job that is left is a better one. We become operators and orchestrators. We manage the agents, keep them in line, stay in the loop, and own the result. It is a little funny, because the oldest complaint in management is that the hard part is the humans. Now we are learning to manage machines too.

AI does what it does best, faster, while you keep the final approval and the responsibility. Your team gets the same deal: hand off the parts that drain you, and spend your energy on the parts you love.

Make an "It sucks that" list

Here is what to actually do. Get a piece of paper. Keep it next to your keyboard this week.

Every time you hit something you dread, write it down, starting with the same two words.

It sucks that:

• …I have to chase the team for status updates.
• …I rebuild the same report every month.
• …I have to dig through call notes to remember what a client said.

Keep adding to it as the week goes. Do not force it. Just catch yourself.

Here is the secret. Every "it sucks that" is a prompt. It is a human telling a machine exactly what they do not want to do. And almost every line on that list is something AI can take off your plate, so you can do better work.

Before you let AI hand that time back, decide what it is for. Write one more line at the bottom. A boundary. Bedtime with the kids. Eight hours of sleep. Friday afternoon, phone only. Because AI will give you the time either way. The only question is whether you spend it on what you love, or quietly refill it with more.

The goal was never to build more at 1am. It is to close the laptop, walk upstairs, read a bedtime story, and know the work still gets done while I am gone.

AI for better work. Not worse lives.

If that is the kind of company you want to build, this is why I write. I am putting the whole belief system in one place at humanaimanifesto.org.

If it lands, join the list. And if you want to see where this is all going, AI for Better Lives is the bigger story. Better work is only the first chapter.