Joe Wilkins, writing in Futurism (a hub that started off a Knight Foundation grant in 2017 and has grown from an infographic on a subreddit into a news source read by millions, according to its website), had a piece called You’ll Snort-Laugh When You Learn How Much AI Actually Added to the US Economy Last Year.
The point: “There is no evidence that AI deployment is either boosting productivity or damaging US employment.”
It circles a single idea: despite massive investment, AI hasn’t meaningfully moved GDP. That observation is now being used, in some cases, to question whether AI is delivering real economic value.
Think about that.
The logic is trying to say AI hasn’t moved GDP. That raises a more important question: what, exactly, are we measuring?
Wilkins is measuring the wrong thing. And worse—he’s measuring it in a place where AI’s impact won’t show up yet.
There was a saying our company lived by: If you can measure it, you can control it. The core problem with trying to measure AI’s impact is this: We’re trying to measure it in a system that wasn’t designed to see it.
We’re Looking for Measurements in the Wrong Place
GDP is a lagging indicator. It tells you what has already scaled—not what is actively changing.
It is primarily affected by consumer consumption, business investment, government spending, and net exports (exports minus imports)—a layered set of forces. Those components are driven by labor quality, capital investment, technology, and interest rates, which together shape economic production.
High-level stuff. The stuff of economic models. But not always the stuff of business reality.
AI is not only having an effect on the economy; it is shaping it in quieter ways. Consider this: AI has not yet scaled into a system-wide productivity engine. It is something else right now. It is a tool being deployed inside companies to change how work gets done.
That distinction matters—both in terms of its visible impact and how it will eventually contribute to measurement. Because what happens inside firms does not immediately show up in GDP. But it does show up in how those firms operate.
And that’s where the real movement is – the measurement that CAN be measured.
What Companies Actually Do With Disruption
There is a pattern to how organizations respond to a new tool like AI—and that response isn’t unique to AI. Companies do not expand first. They contract.
Whenever disruption occurs, companies take that opportunity to “clean house.” They simplify. Compress. Remove dead weight. AI is being used in exactly this way today — not as a growth engine yet, but as a mechanism to:
- eliminate redundant roles
- reduce layered review structures
- accelerate tasks that previously required multiple hands
Processes that once took days are now reduced to hours. In some cases, minutes. The immediate result is not more output. It is the same output, achieved with fewer inputs and less friction.
Speed.
And speed is not something GDP captures cleanly in the short term.
Speed Is the Real Change Agent
The most overlooked effect of AI is not intelligence. It is speed.
Tasks that once required multiple iterations, stakeholders, or days can now be executed by a single individual in a fraction of the time. That does three things immediately:
- Collapses decision cycles
- Reduces the need for validation layers
- Increases throughput per person
None of these effects show up cleanly as “growth.” But they do show up in layoffs, in role compression, and in the constant redefinition of what companies actually need.
This is how organizations change.
The “Cleaning House” Phase
What we are seeing right now is not expansion. It is restructuring. Companies are using AI the way they have used every major disruption: First, to remove what they no longer need.
That includes:
- redundant analysis functions
- duplicated administrative roles
- slow, multi-step processes
This is uncomfortable to say plainly, but it is accurate: AI is reducing the need for certain skill sets.
Think about what Excel® did to manual bookkeeping, manual data tabulation, and complex arithmetic calculation. What GPS did to reading a map. What digital photography did to film developing, specialized lighting adjustment, and darkroom chemical knowledge. Need I go on?
These are not failures. These are structural changes.
Why the Data Look and Are Misleading
At the same time disruption is happening inside companies, capital is flowing heavily into AI infrastructure. Consider the position of Nvidia:
- dominant share of AI accelerators
- overwhelming control of GPU supply
- millions of AI chips deployed into data centers
- a significant portion of global semiconductor capacity tied to AI workloads
This is a capital-intensive buildout. And much of it is global. So while investment is high, the domestic GDP effect is diluted by imports, concentration of gains and uneven adoption across industries. This tells us something important: The buildout is ahead of the application.
Again, this does not mean nothing is happening. It means the activity is not evenly distributed—or immediately visible in aggregate metrics.
The Sequence Matters
What we are seeing now follows an all-too-familiar pattern:
Phase 1 (current):
- Internal efficiency gains
- Role compression
- Speed replacing process
Phase 2:
- Workflow restructuring
- New operating models emerge
Phase 3:
- Output expands
- Market-level productivity becomes visible
GDP gets reflected in Phase 3. We are still in Phase 1.
Calm Down and Let Yourself Be Disrupted
A business associate told me recently they were fearful of AI because it is so disruptive.
I replied: “The only place opportunity gets a chance to emerge is in disruption. Let it happen—and watch for those openings.”
AI is not failing to impact the economy. It is doing what real disruption always does first: Changing how work gets done before changing how much gets produced.
The how before the how much.
Makes complete sense, doesn’t it? Over time, that will translate into measurable growth.
But not before it reshapes the structure of work itself.
Therefore, the real question is not whether AI is contributing to economic growth. The real question is whether we are willing to recognize the form that contribution takes in its early stages.
Right now, that form isn’t expansion. It’s not even acceleration.
It’s compression.
And historically, that is what always comes first.