Today, every business leader in the world is asking this question: How do we AI-ify our business, our products, and our organization?

Some CEOs are drawing a line in the sand. Tobias Lütke from Shopify mandated AI use for all employees and Luis von Ahn at Duolingo announced his intent to phase out contractors with AIs instead. But broad institutional change isn't easy. As this BCG study found last October, 74% of companies are struggling to roll out widespread AI use in a way that adds broad value.

What I'm finding is: Most organizations don’t fail at AI because of the tech. They fail because they haven’t built the cultural habits to learn, adapt, and let humans work differently.

So what's a business leader supposed to do about it? And what does it really mean to build AI fluent orgs?

This year, I decided to see how far I could go on my own by turning a project (see: MuseKat) into an app, and ultimately (hopefully), into a business.

As a one-person team, I knew I would only be able to do this if I completely AI-augmented every part of my workflow, across basically every area of work. Which, essentially, is what I've done. (You can see some slides I made about this process here.)

Essentially, I wanted to see what it would look like, not only by "vibe coding" my way to v1 engineering and product prototypes, but also including "AI in the loop" for every other part of my workflow as well. Here's an example of what some of these non-engineering workflows looks like for me:

As a one-person org, in some ways, I have it easy. I get to learn all about AI, on my own terms, in a way that makes sense to me.

I can try one workflow one week, then scrap it all for a new, better one the next week, without missing a beat. I don't have to wait for colleagues to "catch up" to my erratic experimentation shifts. I don't need to manage personalities and egos when I change my mind about my preferred LLM, my design tool du jour, or even the product direction for what I'm building (which, as I'm finding, can change as quickly as it takes me to vibe prototype a new thing).It's easy for my AI's to bend to my flightiness in tech rollout.

This work mode comes with its benefits. When ChatGPT happens rolls out a major release that supplants the need for us to finish the first task, as a business leader, it's my job to respond to that in real time. That might mean dropping a project, or pivoting completely. It might mean reassigning work, moving deadlines around, or seizing a more opportunity short-term growth channel.

In other words, there's no such thing as "business as usual." Everything is "chaos mode."

This is a stressful way of life. And it'd be a hell of a lot harder to work this way if I were managing humans full time, day in and day out. First of all, I'd need people who could keep up. (Not impossible, but it would take time for both of us to adjust.) Humans also require patience, care, and well, a human touch. In general, people tend to like it when you tell them something on Monday and that thing still holds true on Friday.

Interestingly, when I bring humans into my work at MuseKat, it's always come at the expense of short-term speed and efficiency gains. Humans book meetings three weeks out, then cancel on me the day before. Humans don't respond to my emails right way. They make promises they don't keep. They take time to get work done, then they (still) need to make changes, which means we need to go through another round of edits.

Rather than see humans as a constraint to my hyper-productive work, I've started inverted the question a bit to instead ask, "Where might it be better for me to slow things down? How can humans help me pace myself in a healthier way, or give me more time to pause and reflect, together?"

That's the one-person org conundrum in an AI era. I need humans to slow me down.

Obviously, going all-in on AI as a solo operator is very different from scaling it with other humans in the mix. That’s when things get… slower. And messier. So how do you know where to start?

One of the common tricks often talked about when first teaching people how to use AI systems is this idea of "human in the loop." This means identifying the point where a person steps in to make a decision at a key point in the technology-enabled process. This can refer to simple processes (ie: requiring a human to sign off on AI-generated social media copy), but it can refer to higher stakes situations too, like a doctor stepping into an AI-led conversation during a medical emergency.

When I worked with the Decoded Futures team last year, we spent a lot of time working with nonprofit leaders on creating process maps for different workflows within an organization. At each step in the process (say, you're looking to work with your team on writing a grant application), we would ask the question: Human, AI, or AI-enhanced?

In other words, at what point in this process is it most helpful for humans to be involved, and at what points is it OK to offload that task entirely to an AI? Those AI-enhanced or AI automated steps represent great opportunities to inject collective learning into your company culture.

And that's where the real fun begins.

I don't care what people say, learning is hard.

Whether you're teaching your kid how to ice skate or trying to teach yourself AI, it's not a straightforward path. Learning is full of tough questions, external blockers, and internalized self-doubt.

One of the things I hear a lot about how tricky it is to learn AI is this incessant need to keep learning more, faster, and faster. I don't disagree. It's exhausting (and sometimes scary) to feel this constant pressure to perform. As Tina He wrote in her post, Jevons Paradox:

As individuals become much more empowered, how does one stop oneself from working, that now the opportunity cost of not working becomes almost infinite?

I'd extend this even further. As the opportunity cost of not learning becomes almost infinite, how do we stop? How do we know when to pause, or give ourselves permission to be okay not knowing?

Over the past five years, I've worked in about a dozen different jobs, across half as many industries. I've constantly and repeatedly thrown myself at the hard thing in order to find a way through to the other side. And one thing I've noticed is: When you're in an active learning state, it doesn't always feel good.

You get stuck (more times than you care to admit). You need to willingly delay gratification of something else in order to put in the hard work. You need to take a little bit longer to do it the first time. You need to feel like a failure in order to get over the next hurdle. You need to ask for help.

And if you're in the lucky position of managing people who are going through an active learning state, you need to be willing to sit with them through all of this messy discomfort. Then coach them through all their anxieties, as painful (and slow going) as it might be.

Learning like this doesn't come naturally to all of us. Especially not if you’ve spent years doing the same job, in the same way, with little incentive to change. Or if you work in an industry that rewards coloring inside the lines, where the old methods still work "good enough," and novelty feels like risk.

Maybe over the years, your performance reviews have focused more on compliance (as opposed to creativity). Maybe school taught you how to memorize facts, not to explore possibilities. Either way, if you haven’t built the habit of learning as an adult, the first steps can feel impossible. Trust me, I've been there. Again and again and again...

But not everyone is wired to chase discomfort. Not everyone has the patience or the psychological safety to fail in public, ask for help, or linger in that sticky, murky, in-between stretch zone long enough to break free.

That’s why, as business leaders, people managers, or simply as humans, one of the most powerful things we can do right now is support each other through this massive, messy, collective upskilling moment.

Maybe this is what it really means to have humans in the loop.