Restructure to Survive
Technology is expanding faster than it ever has. The tools being created are incredible. Every week there's a model or an agent doing something that took days, even months, of work previously.
But no one seems to be addressing the elephant in the room. Creating a new technology is one thing. Integrating that technology is another.
How do you integrate these tools into your workflows? How does your team structure need to change to support this new way of working? What has to be removed or added to ensure success?
I realize this isn't the "sexy" part of innovation, which is why nobody wants to talk about it. But these are the questions you need to start asking if you want to come out on top of the AI boom.
Bill Ackman put it simply: "Every CEO in America today is asking how to use AI," noting that the risk of disruption "has gone up dramatically."
There are two problems I'm seeing on a macro level right now across all big companies, relative to the new technology being released daily. The good news is there's one simple, effective solution to both.
Technology vs. The Physical World
The technology industry is known for its constant innovation, which includes multiple large scale pivots and reorgs a year. These companies are highly financially incentivized to create new ways of doing things, because the risk of having outdated technology outweighs the risk of abandoning previous investments. Move fast and break things is their motto.
However, most things in the physical world cannot move in the same manner. Construction is a great example, and it's the one I've spent the last 10 years in, building commercial and mission critical projects across six companies, including Meta.
Construction has inherently been the dinosaur in terms of advancing technology. It's extremely process oriented, which makes sense when you consider that changes (or mistakes) can cost millions in change orders and deteriorate the bottom line.
Now, when you're building something like data centers, you're seeing these two entities forced to work together, and it's quite painful to watch. A new chip comes out every few months. The infrastructure needed to support it takes months of coordination, months of installation, and an ungodly amount of money in rework. These billion dollar projects are being turned upside down last minute, because for once the financial incentive to do so is far greater than the massive cost of the changes.
But the way those changes are being integrated is haphazard and messy, leaving 90% of projects delivered late to customers who need to store their data and train their models. Publicly, roughly half of the US data centers slated to open this year are delayed or cancelled. From the inside, it's worse. There are billion dollar decisions moving through Slack and Teams channels. A ton of risk, a ton of lost time, a ton of mistakes being entertained. It is quite literally the wild west of our generation.
Neither side is wrong. Tech says the chip is ready, and it is. Construction says you can't just drop it in, and they're right. There's no shared language for the gap in between.
What you need is structure with flexibility. Constraints you genuinely cannot break, ensuring proper coordination. This kind of structure is thought to "slow things down," but it actually enables teams to move quicker without taking on the risk of missing something in the details that could cost millions or months to fix, depending on when you catch it.
And this applies to every company, not just the ones building data centers. There's a lot of new tech coming out, but integrating it into the real world is not that simple. It takes planning, and it takes structure. In the same way a CEO might want to rush integration, the extra thought and planning up front is what makes integration actually work. We need to build the structure for how these new systems will communicate with the old ones.
Which brings me to the next issue.
AI & The Exponential Risk of Disruption
Jack Dorsey cut 4,000 people at Block. Forty percent of the company. He and Roelof Botha published the thesis behind it: hierarchy is an obsolete information-routing protocol, so collapse five layers of management into two.
What he did here was monumental, and most companies missed it.
Our old way of structuring businesses is detrimental to the integration of AI.
Let me explain. Most companies have layers of middle management that run work and control output. In the past this was necessary. Now it is slowing down a company's ability to innovate and get to market.
Speed to market is the killer here, because with the new AI tech a solo founder with their laptop alone can create entire software systems and release them by lunchtime. Using the same tech, that release would take six months inside a large company.
Big companies have a lot of leverage, of course. Brand recognition. Service packaging. But that means nothing when you're offering an outdated, overpriced product.
The key here is simple. You must enable the use of AI within your company in order to keep up with the speed of these new AI-backed entrepreneurs.
How?
Most large companies are simply adding AI tools to their employees' toolbox and asking them to use them as much as possible.
This is a weak strategy that is being widely used, and I don't understand why. The problem is not how well employees are integrating. It's what they are building these tools on top of. It's a structural problem.
AI cannot run on chaos. It can't run on handshake deals, tribal knowledge, and six layers of management that exist to translate between each other. There is nothing there to automate.
And the structure it would need doesn't exist. Not in a document, not in a system. It's sitting in your best people's heads, and it always has been. You just never had to write it down before, because they were the ones running the work.
Doesn't a deliberate transition seem more organized and thoughtful than laying people off haphazardly and throwing AI at the remaining employees to figure out while they're burnt out doing the work of their old counterparts?
Ford ran this experiment in reverse. They leaned on AI quality inspection, then brought back roughly 350 engineers in June because the systems weren't catching defects. Their VP said it plainly: AI is only as good as the information you train it on. They got to the right answer. It just cost them the defects and the year to get there.
You must set up the framework first, before trying to integrate AI.
The Solution Is Simple, Not Sexy
Both problems come down to the same gap. Your tech and construction teams don't have a shared standard to work from, and neither does the AI. So build one.
Find your top performers. Not the loudest people. The ones whose projects come in clean.
Document what they do. Not a job description. The real sequence. What triggers the work, what they check, what they escalate, what "done" actually looks like, and all the judgment calls they make without ever noticing they're making them.
Then find the gaps and define them. This is the step everyone skips, and it's the expensive one. Documenting your top performers only captures work that already has an owner. The failures that cost you live in the space between teams, where nobody's process applies and there is no expert to go interview. How does a late-stage change actually get executed? Who approves it, in what order, and what has to be re-checked downstream once they do? Who carries the risk when tech moves and construction has to absorb it? Nobody's head holds that answer, because it was never one person's job. You have to write it from scratch. That's what closes the gap between the two industries, and it stays invisible precisely because there's no one to ask.
Put it in a system people will actually open. Asana, Monday, whatever your teams will touch daily. A perfect process nobody uses is worth less than a decent one everybody follows.
Then layer AI on top. Now there's training data. Now there are boundaries. Now there's a standard to measure against. As the models evolve, you widen their coverage.
The first two steps capture what your people know. The third builds what nobody knew, because it never belonged to anyone in the first place. Put them together and you have structure with flexibility. The documented process is the part you don't break. The flexibility is that anyone can now execute inside it. A new hire. The other side of the project. A model. None of them need the one person who knows everything standing in the room.
And then the transition happens on its own. You hire less technical people to run highly complex projects. You have fewer people overseeing more work. You hire against a defined role instead of a feeling. You onboard in weeks instead of quarters. Your best person taking leave stops being a schedule risk.
And you finally find out which management layers were adding judgment, and which ones were just moving information from one place to another.
Start Before You're Forced To
Do not press the panic button and lay people off without consideration. Don't throw AI at your employees. Future success takes strategic, methodical planning. Take the chaos out of it and create the structure AI needs to succeed.
I truly believe this is the single differentiator that will determine how successful a company is over the next one to two years. If you are not laying the foundation now, you will not be ready to implement the models that are evolving as we speak.
