The layoffs being walked back, the hidden costs, the jobs debate, and the small businesses winning right now all point to the same conclusion. Written with Brennan McDonald.
We spend our days on opposite sides of the same idea.
One of us spent his career in the corporate world, in program, project, and change management inside big companies, including a major bank, and now writes about getting AI to work. The other is not technical, built a couple of tech businesses anyway, and now builds outsourced virtual teams for businesses.
We both love AI. It makes our lives so much easier. It makes things so much more efficient. A lot of the systems we run every day are built on it.
And we keep landing on the same conclusion from different directions.
These tools are only as powerful as the individuals managing them.
That one sentence runs through every argument this space keeps having. Let’s take them one at a time.
The right human is the superpower
Business owners come to one of us every single day and say: can’t I just get AI to do this?
And it’s partially true. AI does play a role. AI can do the tasks.
But underneath the question is a bigger one. Can I just replace the entire team? Can I remove the judgment? Can I remove the thought process that goes into something?
That part is not true, and here’s what we’ve noticed about why.
With all of these tools, there is so much more we can do now. There are so many more things to manage. They’re not creating less work. They’re creating more work, because we can do more.
And when you find the right person, someone you can delegate that system to, that’s when you create a superpower. The AI is great. It does its job. It deals with a lot of tasks. But when you find the right human to tie around the AI you’ve built, the systems you’ve built, that’s when you’re unstoppable.
Finding that right human mostly means someone with a bit of experience. They know when things have gone wrong before. They know what to look for in the quality of the outputs. They know where the judgment calls work and where they don’t.
A CTO one of us knows put it plainly. He uses AI every single day in his work. But take an expert, someone who’s been developing for 20 years, and give them the power of AI, and that’s the right person given the right tools. There’s a difference with someone who knows how to orchestrate and take control of these tools. That’s when you see the benefits.
Remove the experts from the picture and it can be detrimental for an organization.
Some very big companies just proved that in public.
The layoffs are walking back
Klarna laid off 700 people. The founder was all gung-ho on AI: we’re going to implement all these systems, we’re going to get rid of all these people.
And guess what? He walked it back.
When he did it, he realized the human experience was really affected by it. There’s a ton of case studies now showing big companies that tried to go all in on AI, and what they found is that if you do that, you lose the experts who knew the ins and outs of the system.
When companies make these big layoffs, they’re losing a lot of institutional knowledge. They’re losing the people who know what broke last time and why. And when you’re trying to roll out complex workflows, it’s the edge cases, the ones that only pop up every few months or every few years, that break the attempts to just automate everything.
The businesses that were loudest about replacing people are walking those statements back. They still see the value. They also understand now how much the human experience matters.
Gartner expects that by 2027, half of the companies that cut staff because of AI will be hiring those people back, often under new job titles.
And the people are only part of the bill. The rest of it is hiding in the costs.
THE COSTS NOBODY BUDGETS FOR
A lot of people think AI lets you skip steps. You can get a lot of output faster, yes. That doesn’t mean you don’t need judgment. You still need to know exactly what the acceptance criteria are. What does good look like? What does great look like?
And just because you push something to a system doesn’t mean you stop paying for the maintenance, the tuning, the testing, the upgrades, the compliance, the regulatory side.
You can’t lock, load, and forget. It’s constant, continual improvement, and the technology is moving so fast that something that made sense 12 months ago is now completely out of date.
Scale makes it sharper. If you’re running a workflow hundreds of thousands of times a day and the cost of the models changes, how does that work for your business?
Because the prices will change. Right now AI is really cheap, but the costs are subsidized. There are costs we don’t even know exist yet. The whole theory says AI makes everything cheaper, and it can make things more efficient, but there are a lot of hidden costs, and the prices these tools are at right now are not going to stay at this level forever.
Sometimes the hidden costs arrive overnight. A friend of one of us was building everything with AI coding tools at his company. Then the company started going through SOC 2 compliance, and it changed everything. He had to adopt a new system. Everything had to attach to a protection layer. The cost went up by $25,000 overnight, and now he’s scrambling to train the entire team.
Go in anyway. Just do it with your eyes open.
And if it’s complicated for a small business, step inside a big company and the weeds get thick.
INSIDE BIG COMPANIES, THE WEEDS GET THICK
Some big companies can’t use AI for hiring. They can’t use it for anything credit related. They can’t use it for anything risk or compliance related.
If you’ve never worked in those industries, you might be thinking: what on earth is going on there? Aren’t those the most logical places to use it?
It comes down to the human in the loop.
If you’re the regulator of a bank, and someone is the head of anti-money-laundering compliance, the regulator does not care that an AI wrote the report that had a mistake in it. It’s that person’s problem. They have the regulatory obligation. They’re a named officer. In some cases the fine goes to them personally, the bank can’t reimburse them, and the penalties are actual criminal penalties.
When the fine has your name on it, you check the work yourself.
Security stacks on top. A lot of businesses are terrified of rolling out AI because you don’t just need a cybersecurity team that’s good at cybersecurity anymore. You need a cybersecurity team that’s top notch at AI as well. That’s a double skill stack just to keep the business safe, and it barely exists in the market.
The rules are still being written too. In Australia there’s already a push toward regulation of AI in the workplace, possibly even required consultation with workers before AI changes how their jobs run. Every country is going to be different.
Hold all that together and the strange scenes make sense. There are people in financial services who can only use AI on their personal devices at home. At work they’ve got one basic assistant and nothing else, so their experimentation happens after hours. Which is wild, and it’s where parts of the corporate world are right now.
The leaders getting through it fastest give their people the honest version. If we roll these tools out, we’ll be competitive for longer, our cost economics get better, and we won’t be doing panicked layoffs later. As soon as people have that click, that it helps them do things better, faster, and cheaper, they stop fearing it.
Big companies will adapt. But their speed limit is real, and it shapes the jobs question everyone is asking.
THE JOBS QUESTION, WHERE WE DON’T FULLY AGREE
This is the topic with the most fear in it, and the fear is real. One of us makes short videos about AI at work, and the amount of anti-AI hate he has to block in the comments is shocking. The gap between how business people talk about these tools and how much of the workforce feels about them is wide.
The two of us read the question differently, and we’d rather show you the honest split.
One of us is more cautious.
A lot of workflows can replace teams doing work that’s well documented in standard operating procedures with clear acceptance criteria. The most technical, most experienced people in those functions will be fine, because they can bring that experience and judgment to bear. The more junior to mid-level roles are in for a real struggle unless they rapidly fold in AI skills. If you’re a younger person with a specific domain expertise, the most important thing you can do is push AI as fast and hard as you can, so you can chase that skill bar and not get left behind.
He’s also watched companies offshore work for years, and the lesson transfers. What’s the difference between offshoring something and pushing it to a workflow? You still need skilled people onshore. You don’t just send the work away. Someone still has to be responsible, be accountable, understand what’s going on, and deal with crises. That’s a skill set, and it operates at a much higher level. Managing teams of agents and AI workflows is that same emerging skill. And there’s a span of control: beyond a certain number of functions, balls get dropped, and AI still can’t magically clean that up.
The other one of us is bullish, especially on young people.
Now is the time to go out and learn. Back in the day, a developer had to get really good at finding information on Stack Overflow. That was the skill. Now the tools research it and hand it to you in seconds, so someone starting their career needs to learn how to ask the right questions instead. That information is all there, and a curious person can build the skill for the price of a subscription.
Experience can even cut the other way. A friend of his works at a development firm where everyone is an experienced developer, and their biggest challenge is adopting the new systems, because their architecture principles rely on outdated practices. All those previous systems in place make it very difficult to change. A junior carrying nothing can move faster than a veteran carrying everything.
Where we land together is one word.
Race.
Whatever your seat, whatever your level: be racing to learn these systems, to catch up, to close that skill gap, while the tools are cheap and almost nobody is far ahead of you.
And the people racing fastest are not who you’d expect.
THE SMALLEST BUSINESSES WIN THE MOST
One of us has friends who sped up their legal research by an order of magnitude with things they put together themselves. They didn’t pay a vendor an enormous sum for an AI legal research tool.
And no, these aren’t tech people playing lawyer. These are lawyers doing it.
For smaller firms and independent operators, that’s where the opening is. The tools have the power. They have the domain expertise. The gap was only ever the technical skill and the quality control, and that gap is closing. Why pay tens of thousands a year for software that does half of what you can put together with a subscription?
Smaller firms might grow from two or three people to ten, because they can run a bigger book of business profitably, at high quality. And the bigger companies that are too slow, whose policy says we can’t use AI, are going to feel that competitive pressure.
One of us was talking to a client about exactly this. The client had hired a firm to build software for his company, and he didn’t really understand what he’d bought. He didn’t have the power to build it himself eight months ago. Now he’s doing it himself, and he moves a lot faster, because he understands the problem so well.
You’ll start seeing businesses everywhere building their own software. That will be what makes them competitive. They’ll have their own tools nobody else has, because they understand the problem way better than everyone else.
Eighteen months ago, one of us couldn’t build his own website. Now he’s built the website and he’s building his own applications.
The prize is bigger than revenue, though. The work itself gets better.
THE BORING PARTS WERE NEVER THE JOB
Think about fulfillment in your work. The joy in your work.
Doing repetitive tasks day in, day out is the mindless stuff people don’t want to be doing. We know teams where people spend their days just checking off the box, and it’s not enjoyable for them. They’d rather be solving problems and thinking about the role as a whole.
Those tasks can go away, and it’s better for everyone.
Think of a lawyer. Do they want to draft the same repetitive contracts every single day? Does the doctor want to run the same intake questions? No. They want the complex problems. They want time with the patients.
One of us used to build technology in healthcare, and the running joke was the doctor pulling up WebMD on their phone mid-visit. Before that, they left the room to check the big encyclopedias. The value was never the walk to the encyclopedia. The value was the judgment about what the information meant for the patient in the room.
Push away the less interesting, repetitive stuff and people get to focus on the interesting problems, the ones that give you some self-actualization at work.
Honestly, if the two of us never had to look at an email or a spreadsheet again, we’d be happy.
Which leaves the last topic. The one about you.
BE THE MASTER OF YOUR OWN DESTINY
There is so much opportunity for folks right now.
It’s like when the iPhone came out and everyone started building mobile apps. We’re there. Everyone has an opportunity to build something really awesome, to get better at their craft, to learn more, because the information is so readily available.
And right now is the cheapest it’s going to be. There’s so much investment capital in these tools, and every company is trying to lure people onto theirs. It’s cheaper than it’s ever been to build a business and to grow a business.
Even if you’re skeptical, lean in. If you don’t lean in and learn, you can’t make informed decisions about where you take your business.
So whether you work for a company, you’re building something, you’re a solo practitioner, or you’re a professional: take the power into your hands.
Get started. Start using these tools. Start adopting them. You don’t need to understand the complexities. Figure out how to use it in your world.
We’re all masters of our own destiny in this era.
The person who runs the system is the most valuable part of the system. It might as well be you.
What’s one task in your world you’d hand to AI this week?
Tell us below.
We’ll both be in the comments.
