← A Thousand Words
By Sven Fund · July 26, 2026
Diesen Artikel auf Deutsch lesen →
AI does not replace the idea. It shortens the path between an idea and the moment when others can see, read, understand and use it.
Only a few days ago, I would have hired several specialists to produce what is now on my screen: researchers, fact-checkers, editors, designers, programmers and publishers. Today, I formulate an idea, debate the thesis, have counterarguments examined, and receive a publication-ready article, a cleanly typeset Word document and a website that can actually go live. Not someday. In the same workflow.
That is precisely why one can say—rhetorically and with a little exaggeration—“There is no more Science Fiction.” Not because machines can now do everything. They cannot. But because the boundary at which so many of our ideas used to end has disappeared. For decades, the talking computer that not only answers but also researches, writes, designs, programs, tests and carries out tasks across multiple tools was a motif of science fiction. Today, it is a working instrument.
My experience over the past few days is only a small example. In a short time, an idea became a column—fact-based, with linked sources, and delivered as a visually reviewed Word document. It also became a reusable format for future articles. “We” shortened texts, expanded them and shifted their emphasis. Websites could be prepared, reviewed and published. With the help of technology, I optimized costs that I would never have understood—and therefore never have addressed—without it. What matters is not one individual capability. What matters is that many previously separate steps become one continuous dialogue.
The effect is therefore far greater than the sum of the minutes saved. In the past, the path from idea to result was filled with handover costs: writing briefs, finding service providers, exchanging files, answering questions, consolidating corrections and comparing versions. If I had done it myself, I would have been interrupted, had to find my way back in, and perhaps set it aside yet again. Is that the better process? For some texts and tasks, certainly—but for all of them?
Studies now show that this direct form of effectiveness is more than my personal enthusiasm. In an experiment involving professional writing tasks, ChatGPT reduced completion time by 40 percent while increasing assessed quality by 18 percent. In a study of more than 5,000 customer-service employees, generative AI increased productivity by around 14 percent on average, with less experienced employees benefiting most. In a field experiment with Boston Consulting Group consultants, suitable tasks were completed more than 25 percent faster and at significantly higher quality. And in a controlled programming test, developers using GitHub Copilot completed their task 55.8 percent faster.
Figures like these are no guarantee of quality for every task. Researchers rightly speak of a “jagged frontier,” an uneven performance boundary: AI is exceptionally strong in some areas, while in others it makes serious errors that nevertheless sound convincing. That is exactly what I have also learned from working with it. The greatest benefit does not come from delegating blindly. It comes from effective interplay: the human sets the goal, perspective and quality standard. AI expands the options, handles the laborious work, tests alternatives and keeps the process together.
This also changes the role of small teams. Capabilities that once became available only at a certain organizational scale are becoming accessible: market analysis, editorial production, translation, design, software development and the maintenance of multiple publishing channels. One person does not automatically replace an entire organization. But that person can carry an idea much further before capital, staff or external help becomes necessary. This lowers the cost of experimentation—and increases the number of ideas that get a chance at all.
What fascinates me in particular is the new speed between thought and visibility. An idea no longer has to sit in a note for weeks. It can be published the same day. Speed used to be the enemy of quality. With a well-managed AI process, speed and diligence can, for the first time, increase together: sources become clickable, claims verifiable, layouts are rendered before handoff, and errors are corrected through multiple iterations.
Responsibility, of course, remains with the human. A machine does not put my name under an opinion piece. It does not decide which position I want to take. It does not know my readers, my experience or the consequences of publication. The more powerful the tool, the more important judgment, transparency and responsibility become. But that is not an argument against the technology. It is a description of a more demanding human role.
The true potential of AI therefore lies not in replacing creativity but in unleashing it. Less energy goes into changing formats, routine work, searching and coordination. More energy remains for questions, decisions, taste—and courage.
Science fiction has always told stories about machines that expand our possibilities. We have now taken another step along that path: we no longer have to imagine this future. Instead, we have to learn how to use it well.
Sources
- Noy & Zhang (2023): Experimental Evidence on the Productivity Effects of Generative AI
- Brynjolfsson, Li & Raymond (2023): Generative AI at Work, NBER
- Dell’Acqua et al. (2023): Navigating the Jagged Technological Frontier, Harvard Business School
- Peng et al. (2023): The Impact of AI on Developer Productivity, Microsoft Research
New posts by email
Receive new fullstopp publications directly by email. Your subscription is completed through a confirmation email.