Your people are already using it. The question is whether you’re leading it or following it.
Somewhere in your organization right now, someone is producing work with AI tools you didn’t authorize, can’t see, and won’t evaluate until it either becomes indispensable or causes a problem. That someone might be a developer, an analyst, a clinician, an underwriter, or an engineer on your plant floor. The tool differs. The dynamic is identical, and it’s already playing out in virtually every organization, in every industry, right now.
In manufacturing, engineers are using AI to generate design specs and quality documentation that looks authoritative, whether or not it’s accurate. In healthcare, clinicians are using it to draft clinical documentation and compliance reports, building a liability surface faster than anyone is tracking. In financial services, analysts are building models and client memos with AI assistance, and the confidence in the output often exceeds the quality of the analysis underneath it. The tools are consumer-grade accessible. Your people don’t need permission to use them, and they’re not waiting for you to catch up.
We’ve seen this movie before
When electric motors became commercially available in the late 1800s, most manufacturers did the natural thing: they replaced the steam engine with an electric motor and kept everything else the same. Same factory layout. Same belt-and-shaft system running from one central motor to every machine on the floor. Same organizational structure. New power source.
It didn’t work. Productivity barely moved. For decades, the electric motor sat in factories without producing the transformation everyone expected.
The manufacturers who eventually won didn’t just electrify. They redesigned. They realized that distributed power, the ability to put a motor exactly where the work was, made the centralized factory floor obsolete, and rebuilt their layouts, decision-making, and organization around it.
Economist Paul David documented this pattern: it took roughly 30 years from widespread electricity availability to measurable productivity gains in manufacturing, not because the technology was slow, but because organizations were.[^1]
We’re in that same phase with AI right now. Most organizations are using it to do existing things slightly faster; the redesign hasn’t happened yet, and whoever figures it out first won’t just be more productive. They’ll be operating at a fundamentally different level than everyone still running the old layout.
Why this technology is riskier than the last one
AI has a specific danger earlier technologies didn’t: it produces output that looks professional even when it’s structurally unsound. A financial analyst who makes a calculation mistake can usually find it when the numbers don’t add up. AI-generated work product doesn’t wobble visibly. A hallucinated financial model looks like a real one. A flawed clinical summary reads like a competent one. A security-less software system functions exactly like a secure one, right up until it doesn’t.
That’s the Judgment Gap: the distance between what AI can produce and what professional judgment can evaluate, validate, and stand behind. AI democratizes production. It does not democratize judgment.
The gap is widened by a confidence problem running in both directions: people tend to overestimate AI’s reliability because the output looks authoritative, and organizations reviewing it often can’t tell what was AI-generated from what was professionally validated.
There’s also a scale problem. Unlike most mistakes, AI-assisted mistakes propagate. One flawed template, replicated across an organization before anyone catches it, compounds with every copy, and the cost of fixing it later is far higher than doing it right the first time.
What the winners are doing differently
The historical record on technology transitions is remarkably consistent. Walmart was not the first retailer to use computers. Amazon was not the first online store. Toyota did not invent automation. In each case, the eventual winner wasn’t the first to access the technology; it was the organization that built the most disciplined system around it.
Toyota’s experience is instructive. Through the 1980s and 1990s, General Motors had more industrial robots than Toyota and worse outcomes by nearly every measure of quality and efficiency. GM automated its existing broken processes. Toyota redesigned the process first, developing what became the Toyota Production System, then applied automation to the redesigned process.
Capital One tells a similar story. In the early 2000s, it made a deliberate decision to become a technology company that happened to issue credit cards, rather than a credit card company that used technology, and rebuilt its talent model, structure, and risk management around that identity shift. It came through multiple cycles of disruption more resilient than competitors with more legacy assets and less organizational clarity.
The first-mover advantage in AI is largely a myth. The intentional-mover advantage is very real. In each of these cases, leadership was willing to change what the organization did, not just add technology on top of existing processes. The technology was the easy part. The organizational clarity about what it made possible, and what it required in return, was the hard part.
What intentional looks like
Closing the Judgment Gap isn’t about slowing AI adoption. Organizations that over-restrict AI tools will just have people using them anyway, less visibly, with less oversight. Restriction without engagement isn’t a strategy; it’s an invitation to shadow systems. The organizations building durable advantage around AI share three habits.
Visibility. They know what AI-assisted work is being produced, by whom, and where it’s being used. This isn’t surveillance; it’s the basic hygiene of knowing what tools are in use and what the exposure looks like. Most organizations have none of this. The tools are in use; the inventory doesn’t exist.
Tiering. They have clear criteria for what moves forward on individual judgment versus what needs professional review before it becomes load-bearing. A first-draft email carries different risk than an AI-assisted actuarial model or clinical protocol. Intentional organizations have defined where those lines sit. Most haven’t.
Augmentation over substitution. They treat AI output as a first draft that enters a professional process, not a final product that exits one. This gets harder to sustain as AI output improves and the temptation to remove the professional layer grows. The organizations that keep their edge will use AI to make judgment faster and better, not let it replace judgment and discover the consequences later.
None of this requires prohibiting AI or a major technology investment. It requires leadership decisions about how the organization operates, which is exactly why it’s rare, and exactly why it’s valuable.
The window is open now
The tools are already on your floor. Your people are already using them. The organizations that close the Judgment Gap deliberately won’t just outperform competitors over the next few years, they’ll define what their industries look like for the decade beyond that. The ones that wait, for the technology to mature or for a competitor to move first, will find themselves running belt-and-shaft systems while the redesigned factories come online around them. The gap, once it opens, is very difficult to close.
Want help closing the gap in your organization? TMF has spent more than 45 years helping organizations in healthcare, manufacturing, and financial services navigate exactly this kind of technology transition. Contact us to talk through what visibility and tiering could look like for your teams.
[^1]: Paul A. David, “The Dynamo and the Computer: An Historical Perspective on the Modern Productivity Paradox,” American Economic Review 80, no. 2 (May 1990): 355–361. Read the paper (PDF), or see this accessible interview with David on the parallel to AI: AEI, “The Dynamo, the Computer, and ChatGPT: Explaining Today’s Productivity Paradox”.
