The Factories That Got Electrification Wrong
When electric motors became commercially available in the early 1900s, factory owners faced a decision that probably felt obvious at the time. They had a steam turbine. Now there was a better power source. So they swapped one for the other, kept the building exactly as it was, and waited for the productivity gains everyone was promising.
Costs went up.
Not eventually down immediately up. The electricity was more expensive than coal and steam. The capital cost of installing the equipment was real. And the output? Identical. Same building, same layout, same central shaft running belts to every station in the building. The whole structure had been architecturally designed around one power source. You can’t just replace the engine and expect the factory to work differently.
The productivity revolution everyone associates with electrification the gains that showed up in economic data and reshaped American manufacturing took twenty to forty years to materialize. And it only arrived from the companies that did something most factory owners were deeply reluctant to do: tear the building down and rebuild it entirely around what electric power actually made possible.
Distributed motors. Independent stations. Horizontal layouts instead of vertical ones. Flexible production lines that could be reconfigured without shutting down the whole operation. The gains weren’t in the motor. They were in what the motor allowed you to design from scratch.
I’ve been thinking about this a lot lately because it’s exactly where most organizations are right now with AI.
The “Plug It In” Phase Most Companies Are Stuck In
The conversation that surfaced this framing happened on a recent episode of the podcast with Matt Graham, who put it simply: most businesses got access to AI and did the obvious thing. They plugged it into whatever they were already doing.
A customer service team added an AI chat layer to an existing ticket workflow. A marketing team started using AI to draft content that still goes through the same five-person approval chain. A sales team got access to AI tools and used them to do faster versions of the exact same manual research they were already doing.
All of that produces something. But what it tends to produce is higher operating costs (the tools aren’t free), modest efficiency gains on individual tasks, and a slowly growing sense of frustration that AI isn’t delivering the transformation everyone said it would.
This is the factory manager staring at an electric motor bolted to the old steam infrastructure, wondering why the output charts look the same.
The issue isn’t the tool. The issue is that the structure it’s plugged into was designed around the constraints of the previous era. Every workflow, every approval chain, every job description, every budget line all of it was built to make sense in a world before this technology existed. Plugging a new power source into that structure doesn’t change what the structure produces. It just changes what it costs to run it.
Why do most AI investments fail to deliver ROI? Not because the technology doesn’t work. Because organizations implement AI tools without redesigning the processes and structures those tools are plugged into. The same pattern played out during factory electrification: swapping the power source without rebuilding the floor produced higher costs and flat output. The productivity gains only came when companies rebuilt from scratch around what the new technology actually made possible.
What “Rebuilding the Factory” Actually Means for a Business Leader
Here’s where the electrification analogy earns its keep because it also clarifies the decision in front of leaders honestly, without oversimplifying it.
In the conversation with Matt, we laid out three real options. Not as a framework someone drew on a whiteboard, but as the actual choices that exist:
The first is to do nothing meaningful. Keep the current structure, plug in AI where it’s easy and low-risk, and essentially manage the company toward a terminal value. This sounds cynical, but it’s a legitimate business decision if made with open eyes. The yellow pages companies after the internet are a useful example investors bought them cheap, fully aware they were declining assets, made money on the cash flows, and moved on. Some businesses are in that position. The mistake is doing it accidentally rather than intentionally.
The second is to start from scratch. If you’re building something new, or if your competitive environment is changing fast enough that the old structure is becoming a liability, the cleanest path is to make every design decision AI-forward from day one. The example I gave on the podcast: instead of hiring a $300,000-a-year marketing team, spend that same $300,000 once to build the automations that do the same work. The cost comparison changes entirely when you’re not carrying the weight of how things already work.
The third and the hardest is to rebuild while the factory is still running. This is the situation most established businesses are actually in. You have processes that work, people who depend on them, a culture built around how things get done, and now you’re supposed to fundamentally redesign all of it while keeping the lights on. That’s not a technology project. That’s a change management project that happens to involve technology.
As I put it on the podcast: it’s easy to be glib and say “just tear down the factory and rebuild it.” But when you’re above twenty or fifty people, when you have established budget flows and organizational patterns and a company that’s profitable and growing at ten percent a year, walking in and saying “let’s redesign everything” is an enormous ask. Understanding why leaders are reluctant to do it is not the same as saying they’re wrong to be reluctant.
The point isn’t that everyone should blow up their organization. The point is that the companies who are already ahead meaningfully ahead, not just-bought-some-tools ahead started having the rebuild conversation earlier. Not because they had more courage or better AI tools. Because they started developing the organizational reflexes for this kind of change before it felt urgent.
Why the Companies That Started Three Years Ago Are Already Hard to Catch
This is the part of the conversation that I think gets undersold in most AI discussions, because it doesn’t fit neatly into a tools comparison or a use case breakdown.
The real competitive advantage being built right now isn’t in the AI stack. It’s in the organizational muscle memory of working with AI the accumulated experience of running experiments, hitting failures, adapting, and building instincts for what actually works versus what sounds good in a pitch deck.
You can’t buy that at any price. You can’t hire it either, at least not easily. An agent orchestrator someone who can manage and direct multiple AI agents working in parallel on complex tasks doesn’t really exist as a job title you can post for yet. The people who will fill that role are being built right now inside organizations that started experimenting early, through a process of learning that’s measured in years, not weeks.
The practical implication is this: the gap between companies that have been building organizational AI capability and companies that are just starting is growing faster than most people realize. Not because the tools are harder to access they’re easier to access than ever. But because the learning curve isn’t about the tools. It’s about developing the judgment to know when a process needs to be rebuilt versus patched, when an AI system is working versus burning tokens in circles, when to trust the output and when to push back.
That judgment is experiential. It accretes slowly. And the companies that have three years of it are not in the same position as the companies starting today, regardless of what either of them spends on software.
If you’re a leader trying to figure out where your organization actually stands, the most useful question isn’t “what AI tools should we be using?” It’s closer to: “What would we build differently if we were starting today, and what’s stopping us from building it that way now?”
The answer to that second part is almost always the real work.
If you want to know where your organization actually sits not the optimistic version, the honest one the AI Readiness Assessment at launchpad.ascendlabs.ai is a good place to start. It’s free and it’s built around the questions that actually matter.
And if you’re ready for a direct conversation about what rebuilding looks like for your specific situation, you can book time at tidycal.com/kevinwilliams. No pitch just an honest look at where you are and what the path forward actually requires.
