A New Time Scale

AI is changing more than tools and markets. It is changing the unit of time in which organizations must learn – and above all act. For scholarly publishers, speed is becoming a strategic capability in its own right.

From Decades to Months

In scholarly publishing, people have learned – and grown rather comfortable with – thinking in long arcs. The first electronic journals appeared in the 1990s. Crossref and the DOI became central infrastructure around the turn of the millennium. The Budapest Open Access Initiative articulated a goal in 2002 that has still not been fully realized. Plan S was announced in 2018 and took effect for many funders in 2021. Even where direction and benefits were broadly uncontested, change took years or decades – an effect many innovators in start-ups have learned the hard way.

That was not merely inertia. Scholarly publishing is a dense system of researchers, publishers, libraries, learned societies, funders and political institutions. Quality, trust and sustainability cannot be updated like an app. But this explanation can also become a habit. Anyone who always thinks in decades may eventually fail to notice that the unit of time outside has changed.

AI Does Not Evolve in Decades. It Evolves in Months.

It has become a truism: AI accelerates dramatically. ChatGPT reached a mass audience within months; by July 2025, OpenAI reported more than 700 million active users – every week. The Stanford AI Index reports that the cost of performance at the level of GPT-3.5 fell by more than a factor of 280 between November 2022 and October 2024. What was expensive, slow or reserved for specialists yesterday can appear tomorrow as a standard feature in a tool everyone already uses.

Adoption inside companies also leaps over conventional implementation plans. Microsoft and LinkedIn reported in 2024 that 75 percent of knowledge workers used AI at work; 46 percent of those users had started only within the previous six months. McKinsey found that the share of organizations regularly using generative AI rose to 65 percent within ten months – nearly doubling. People, including those inside companies, do not wait for a central strategy to be finished. They simply begin using tools that they hope will make their work easier.

That is both the opportunity and the risk. An organization may still officially be ‘evaluating’ AI while its employees are already summarizing texts, analyzing data, writing code or preparing customer emails. Adoption then takes place without shared standards, learning loops or protection for sensitive data. Slowness does not prevent change. It merely makes it invisible.

Speed, therefore, does not mean chasing a new model every week. It means shortening the organization’s learning time.

Learning Instead of Implementing

For publishers, this is very concrete. Within a few weeks they can test whether AI improves metadata, adapts abstracts for different audiences, structures rights information, prepares accessible image descriptions or makes large backlists easier to discover. Editorial teams can examine where they gain time in research, language editing or formal manuscript checks. Sales and marketing can understand audiences and usage patterns differently. Not every experiment will work. But even a failed experiment produces knowledge – provided it is small, measurable and assessed honestly.

The old reflex would be to write a strategy first, then approve a budget, select a system and finally roll out training. That no longer works with AI, and it frightens many decision-makers.

The alternative is not uncontrolled activism. It is a different cadence: create safe workspaces, select real tasks, review the results, refine the rules and share what has been learned immediately.

From the Top to the Front Line

Speed is not the job of a single department. A small AI group can offer direction, but it cannot replace the experience of the whole organization. The best use cases are often found at the front line: among people who correct metadata, answer author queries, check invoices or hunt for production errors every day. They know where time is lost and which exception will defeat every seemingly elegant automation.

Transformation works no better without the top. My professional experience is unequivocal: leaders must use this new category of tools themselves. Not occasionally in a demo, but in their own working lives – when preparing a meeting, challenging an analysis, structuring a text, planning and controlling budgets, drafting contracts or exploring a new market. Only those who do the work themselves understand both productivity and limits. They ask better questions, recognize poor results faster and can credibly decide where people must remain accountable.

The task of leadership, then, is not to ‘implement’ AI. It is to organize a system of learning: from the top with direction, resources and example; from the front line with experience, criticism and concrete cases. Between them, it needs short cycles, clear accountability and the freedom to fail and stop experiments. Speed comes not from pressure, but from reducing the distance between observation, decision and application.

Care Is Not Slowness

Scholarly publishing does not have to abandon its values. On the contrary: quality, traceability and trust become more important when production gets cheaper and convincing errors become easier to create. But care must not be confused with slowness. A six-month decision process is not automatically more thorough than six well-documented experiments over the same period.

The decisive question is therefore not whether an organization already has the right AI strategy. It is how quickly it can learn reliably. Models become better, smaller and cheaper while companies are still seeking advice. Anyone who plans in years what others test in weeks will lose not because of a wrong forecast, but because of slowness and the lack of experience that follows from it.

In the past, scholarly publishing could wait for change, translate it and fit it into familiar structures. This time, the pace itself may be the change. An organization capable of responding does more than work with AI. It becomes more agile – from the top to the front line.