The difficult transition we are going through is, in fact, a problem about the way we write.

At Oakland Tech Week on September 28, Federal Reserve Governor Lisa D. Cook gave a speech in which she once again, for the third time this year, cautioned that the labor market is entering what she describes as a difficult transition due to AI. She stated that the productivity gains intended to counteract the difficulties will come within the next few years. However, they will not arrive fast enough to relieve the pressure in the short term.

In 1987, Robert Solow noted that although the computer age was apparent in all sorts of places, it was not visible in the productivity figures. In 2026, Apollo's chief economist, Torsten Slok, put it this way: "AI is all around us, but it does not appear in the incoming macroeconomic data." As a result, Goldman Sachs and JPMorgan have reduced their 2026 productivity forecast for the US from 2.5% to 1.8%. This situation has continued for forty years: new technologies emerge, large investments follow, but organizations fail to adapt immediately, so the benefits are only seen after a delay. In the 1990s, it took approximately ten years for organizations to catch up.

Since Cook’s speech in September, I keep asking myself: whose decade is this, really? The simplest solution is the next generation of workers. They are about to enter a job market that AI has gradually altered. A study carried out at Stanford by Erik Brynjolfsson, based on ADP payroll data, showed that early-career employees aged 22-25 in positions where AI is used experienced a 16% decrease in employment compared with experienced workers, whose employment remained unchanged. According to Goldman Sachs economist Joseph Briggs, the unemployment rate among young technology workers has increased by three percentage points since the beginning of this year. Cook has identified the entry-level jobs most at risk: coders, back-office positions, and call centers.

This is what they are going through at the moment.

A study by Pierre-Daniel Sarte and Jack Taylor at the Richmond Fed reveals a different aspect of the situation. The productivity benefits of new technology typically appear when today's younger workers reach ages 35 to 44, which is also when productivity peaks in the US workforce. A person in their twenties who has now had a junior position taken away by AI will not see the benefits of the Solow decade until they are in their mid-thirties. If the economy continues to follow the pattern of the 1990s, then it will be they who lead the productivity surge.

That therefore makes the tough decade that lies ahead mostly not their problem but ours to solve.

I lead editorial strategy for the Office of the CTO at an IT consulting firm. My job is to make sure the firm puts into writing what it truly knows, not just what it wants to sell or what sounds appealing. I focus on ensuring a real person in the company can stand by what we write, with evidence, even under tough questioning.

This discipline matters more than ever because the main challenge in producing content has changed.

In 2022, the senior staff had to spend weeks putting together a strategy document, with the document itself being the scarce resource. Today, an AI model can produce a convincing one in under a minute. What is now scarce is not the document but whether anyone in the organization has really thought it through, can defend it, and will stand by it.

Terence Tao, a mathematician, published an essay in August titled "Mathematics in the age of AI". He states that AI systems are now nearly capable of producing research-grade mathematical proofs. During the twentieth century, the mathematical community developed formal foundations so that proofs could be checked. What was previously only implied is now obvious: producing a proof is only one part of the task. The proof also has to be explained, taught, understood, published, and incorporated into what the community knows. If no human can explain the proof at an expert level, then it should be regarded as incomplete, even if it has been properly verified.

The business lesson here is clear.

A strategy document is incomplete if no one in the organization can defend its argument, point to its evidence, or expand on it when challenged. A market analysis is incomplete if no one has read it fully and can explain its conclusions without looking it up. A position paper is incomplete if the person named on it cannot explain why they chose one position over another. Nobody owns it. It joins the stack of AI-generated artifacts that nobody fully read. That is just noise pretending to be real work. The Solow Paradox decade will not work out for any organization that relies on noise instead of real thinking.

Look at what is happening in most companies today. An analyst uses an AI model to create a market study. A senior leader asks another AI model to summarize it. A board slide pulls out three key points. A decision is made based on those points, committing resources. At no stage has a person read the full study, understood the evidence, or tested the argument against other viewpoints. The process is finished, but the real thinking is missing.

Then the strategy memo that follows is also written by an LLM. Next, the strategy memo is also written, read, and summarized by AI models. The people who are supposed to defend the strategy cannot do so because they have never really understood it. The critics cannot challenge it because they have not read it either. The organization moves quickly through a series of documents that no one has truly thought about.etical. It is what Sam Ferrise, our CTO, called this week the gap between the scoreboard that says "AI-first enterprise" and the measured outcome that MIT clocked at 95% zero P&L impact.

That gap exists because real thinking is missing.

When I publish anything on behalf of Trinetix or assist a client with a publication under their name, I ask myself three questions.

First of all, if a real person within the organization cannot defend each and every claim in this document and refer to the evidence, even when closely questioned, then we will not publish it; instead, we will either remove the claim or carry out the work needed to support it.

Second, has each quote been accurately reported, each source been verified, and every figure been traced to a published source or a clearly labeled internal measurement? If not, it is merely marketing, not genuine thinking.

Third: When someone reads this document in two years, will they still be able to detect a clear human perspective or will it instead become part of the background of AI-generated content? If it does become part of that background, then we will rewrite it or choose not to publish it.

Although these might appear to be questions about writing, they are, in fact, questions about organizational memory, because they determine whether what we publish today will form the basis for future decisions or simply add to the pile that the next generation has to sift through to find something useful.

There is a clear effect of the difficult transition Cook speaks about on the job market: the number of entry-level jobs is decreasing, young workers are being pushed aside, and those now entering the workforce are facing a difficult decade.

There is also a less obvious change taking place: it concerns knowledge. Over the next ten years, companies are going to produce a great deal more content than actual thinking.

Although the number of documents will increase, the quality of thought will not keep pace. By the end of the decade, organizations that had maintained strong editorial standards will have accumulated a body of reliable material for the next generation to make use of, while those that had not will be left with a decade's worth of AI-produced documents, each one satisfactory enough to pass inspection but with no real human actually responsible for them. If it ever becomes necessary, the following generation can begin all over again. At that point they will be in their mid-thirties, an age at which productivity reaches its peak as described by the Richmond Fed, and will have ten years of experience with AI. They will be in a position to manage.

What they receive from us depends on whether we give them clear guidance or confusion.

The most overlooked investment of this decade will be preparing documents that a real person can defend, are based on actual sources, and state a clear position. Writing itself is not rare, by any means. The scarcity actually lies in the thinking that goes into the writing, and it is this thinking that will account for any future gains from the Solow Paradox.

The technological and labor aspects of the transition that Cook describes are ones that economists will discuss for years; there is also a subtler aspect that editors, communicators, strategists, and consultants will have to deal with on a daily basis over the coming decade.