# A Broader View AI may make work disappear. But it can also expand the world of work that is possible. The difference lies in what companies choose to look for. - Forfatter: Mikkel Freltoft Krogsholm - Type: Essay - Udgivet: 2026-07-11 - Opdateret: 2026-07-11 - Sprog: en - Emner: ai, work, organization, expertise, future - Kanonisk URL: https://mikkelkrogsholm.dk/en/articles/stoerre-udsyn/ --- Eighteen months ago, I wrote an article called [When Work Disappears](/en/articles/naar-arbejdet-forsvinder/). The thesis was not hard to grasp. As machines can perform more and more cognitive, physical and relational work better and more cheaply than people can, companies will need fewer people. We can debate the pace, but the direction seems fairly obvious. I still think that may be true. I just no longer think it is the only possible future. A new study from Ramp and Revelio Labs tracked 21,559 American companies and linked their actual purchases of AI tools to changes in their headcounts. The result is interesting. The companies investing most intensively in AI did not reduce their headcounts. They increased them by around ten percent in the first two years after adoption. The number of entry-level employees grew even more. That does not prove that AI creates jobs. These companies were not average. They were already larger, more technical and growing faster than companies that did not adopt AI. The researchers themselves are cautious about causal claims. But the finding made me see a possibility I did not see as clearly when I wrote the first article. Perhaps AI is not just a technology that makes existing work smaller. Perhaps it expands the world of work that is possible. ## The world at your fingertips The standard business case for AI almost always begins with efficiency. How many hours can we save? How many enquiries can the agent answer? How much faster can the developer write code? How many employees can do the same work as before? They are sensible questions. They are also very small ones. They assume that the company's world is already known. The products are given. The markets are given. The tasks are given. AI simply has to help us move faster through the world we already have. But what if the greatest gain is not speed? What if it is perspective? Today, I can enter fields I have not been trained in with a speed and depth that would have been unthinkable only a few years ago. I can ask a question about organisational design and follow it into psychology, economics, law, computer science and philosophy. I can get help finding the concepts, research, counterarguments and connections between them. That does not make me a psychologist, economist, lawyer, computer scientist or philosopher. It enables me to see further from where I stand. *The world at your fingertips* has been an advertising phrase since the internet first appeared. But the internet mostly placed the world before us as an immense library. We still had to know the title of the book, find the right shelf, understand the specialist language and know which connections we were looking for. AI makes the library conversational. That does not mean everything it says is true. It means far more parts of the world become accessible as questions. That is a broader view. ## The expert's problem It sounds like an obvious gain. It is not necessarily one. For a broader view requires looking beyond one's own knowledge and expertise. And expertise is rarely just knowledge. It is also identity, status and the right to decide which questions are relevant. The skilled specialist has spent decades learning their field. That produces real and valuable judgement. It also creates a natural inclination to see the world through the concepts, methods and solutions the field already has. If you are very good at using a hammer, the world does not merely become full of nails. You also become the person the organisation asks when something looks like a nail. Then a machine arrives that can search across thousands of professional domains in seconds. It does not have the expert's experience. Nor does it have the expert's loyalty to one particular way of framing the question. That can feel like a threat to authority. Not because the machine necessarily knows better, but because it makes visible where the expert's field of vision ends. The same is true of organisations. An institution does not see the whole world. It sees what its disciplines, procedures, budgets and time allow it to investigate. At some point, it has to stop. That is not a mistake. Endless investigation is not an option, and standard procedures exist in part to protect us from wishful thinking, arbitrary impulses and dangerous special treatment. The problem arises when the limit of the investigation is presented as a limit in reality. “We have not found another option” and “there is no other option” sound almost the same. They are two very different claims. AI cannot tell us what is true simply because it can search more widely. It can find plausible dead ends faster than any human being. But it can make it cheaper to ask: What have we not seen? What other field has encountered a similar problem? What assumption does our conclusion rest on? What would have to be true for us to be wrong? Questions like that can hurt the ego. The alternative is worse. It is to have the world at your fingertips and still touch only what you already know. ## Schrödinger's company That brings me back to work. AI can unquestionably be used to reduce the need for people. If a company produces the same things, for the same customers, in the same markets, but can do the work with fewer employees, the economic logic will be hard to resist. That was the logic I followed in [When Work Disappears](/en/articles/naar-arbejdet-forsvinder/). It has not become wrong. But it assumes that the company's world stands still. If AI instead makes it possible to develop products the company previously lacked the capacity to build, serve customers it could not previously reach, investigate markets it did not understand, and solve problems that were once too expensive to take seriously, something else happens. Efficiency removes work. A broader view discovers work. Both can happen at once. The company is therefore faced with something like Schrödinger's cat. Inside the box, AI is both a cost-cutting exercise and an engine of growth. Only when leadership opens the box through its decisions do we find out which company the technology is creating. Of course, this is not random. One path is easier to calculate. If an agent can reduce processing time by 40 percent, the gain can go into a spreadsheet. If the people freed up can help the company discover a market it does not yet know, the number is much less certain. The saving can be budgeted. The possibility has to be imagined. That is why we risk using a technology with an immense field of vision to optimise what is already visible. The companies in the Ramp and Revelio Labs study that invested only a little in AI saw no statistically reliable change in headcount. Growth was concentrated among the intensive adopters. There may be many explanations for this, and the study does not tell us which mechanism drives the result. My thought is that the difference may not be only about how much AI they have bought. It may be about whether they have made the technology part of the company's way of seeing. A chat subscription can make an employee faster. A company that changes its workflows, its access to knowledge and its ability to act on new possibilities can grow larger. ## The new scarcity A broader view does not eliminate scarcity. It moves it. When it becomes cheap to find information, formulate hypotheses and imagine new possibilities, it becomes more costly to choose between them. The company does not necessarily lack ideas anymore. It lacks the judgement, courage and capacity to carry the right ideas into reality. It is tempting to think that AI can choose for us too. But the choice is rarely only about which opportunity has the highest expected value. It is about the kind of company we want to build, the kind of people we want to be, and the risk we are willing to bear. AI can expand the map. It cannot decide where we want to go. That is also why expertise does not become worthless. On the contrary. As the number of plausible paths grows, experience and judgement become more valuable. But the expert's role changes. The expert must no longer protect their field from questions from outside. The expert must help decide which of the new connections actually hold. That requires another kind of professional confidence. Not the confidence of knowing the most, but the confidence to investigate something one does not already understand. I do not know whether intensive AI adoption will create more jobs across the whole economy. The study concerns early, often technical American companies. Other companies may choose differently, and competition may still end up eliminating far more work than it creates. Schrödinger's cat is still in the box. But I now see a possibility I did not give enough room in my earlier thinking. We have talked about AI as though its most important task were to make work smaller. Perhaps its greatest significance will be that it makes the world larger. --- ## Sources and further reading - Ara Kharazian, Lisa Simon and Ryan Stevens: [A New Look at AI's Impact on Jobs: Firm-Level AI Spending and Workforce Adjustment](https://ramp.com/data/ai-jobs-impact/paper), Ramp/Revelio Labs, 2026. - [When Work Disappears](/en/articles/naar-arbejdet-forsvinder/) – the earlier text this essay both challenges and builds on. - [The Knowledge Explosion](/en/articles/videnseksplosion/) – on what happens when knowledge begins to produce itself.