The AI Productivity Paradox: Why 70% Adoption is Yielding 0% Growth
The recent headlines surrounding AI’s economic impact have been, at best, underwhelming. According to a Goldman Sachs report, AI’s contribution to US economic growth last year was effectively zero. It’s expected that nearly 3 quarters of a trillion dollars will be invested in AI this year alone, so surely that can’t be right?
No impact?!
Even more startling is a recent study from the National Bureau of Economic Research (NBER) which found that while 70% of firms are now using AI in some capacity, a staggering 80% of them have seen absolutely no impact on their employment or productivity levels.
This raises a critical question for every business leader: Are we witnessing a massive failure of technology, or a failure of measurement?
The Modern Solow Paradox
In the late 1980s, economist Robert Solow famously remarked, "You can see the computer age everywhere but in the productivity statistics." We are currently living through "Solow Paradox 2.0."
The disconnect between 70% adoption and 0% productivity growth is a big challenge. If everyone is using it, why isn't it working? Or is it working, and we’re just not measuring the impact correctly?
1. The Implementation Lag
Productivity doesn't move in a straight line. When a company adopts a new technology, productivity often dips or plateaus initially. This is because firms are spending time, money, and human capital on restructuring, training, and trial-and-error. These are intangible investments. We're currently in the trough of the Productivity J-Curve, where the costs are visible but the gains are still compounding beneath the surface.
2. The Measurement Gap
Traditional productivity metrics are designed for an industrial age, they measure output per hour. But AI often improves the quality of work or saves "soft" time that doesn't immediately translate into a higher volume of sales.
If a lawyer uses AI to research a brief in two hours instead of ten, but still charges the same flat fee, that productivity gain is invisible to GDP. If a marketer uses AI to create ten versions of an ad instead of one, and the quality is higher, the output is still just one campaign with one budget. We are likely measuring the wrong things.
3. Playing vs. Integrating
There is a massive difference between a firm where 70% of employees are occasionally using ChatGPT to draft emails and a firm that has fundamentally re-engineered its digital assembly line around AI. Most firms are currently playing with AI, adding it as a layer on top of old processes rather than using it to build new ones.
Focus on Systems, Not Stats
We believe the zero impact stat is more than likelyt a temporary distraction. The measurement gap is real, but so is the implementation gap.
If you are part of the 80% seeing no impact, it’s likely because you’re treating AI as a plug-and-play tool. Real productivity gains require systemic change. You have to change how you work, not just what you work with.
Don't wait for the national productivity stats to tell you AI is working. Define your own internal metrics, time to market, employee satisfaction, or quality of output, and measure the things the economists are currently missing.
