The Creative Process Deserves Better

A creative team adopts a new tool to save time. Within weeks, it has three dashboards, six notification settings, an inbox nobody trusts, and a glowing button labeled Generate. The software is ready to write the campaign. Yet, a human is still required to copy the due date into two project-management systems.

This is progress?

Modern creative operations have settled into a peculiar bargain: we preserve the chores and automate the craft.

Writers, designers, editors, and producers spend their mornings locating files, reconciling feedback, updating status fields, renaming versions, and attending meetings about the meetings required to move the work forward. Then, when the actual creative problem finally appears, a machine offers to solve it before anyone has thought about it for very long.

The problem is not that creative work contains friction. The problem is that we have become remarkably bad at distinguishing between the kinds that drain people and the kinds that develop them.

Some resistance contributes nothing to the work. Searching three platforms for the approved logo does not improve a design. Copying comments from email into a review system does not sharpen an edit. Reformatting the same information for several dashboards does not reveal a hidden truth about the audience.

This is operational friction: effort spent managing the conditions around the work rather than advancing the work itself.

It carries a cost even when the task gets completed. In one controlled workplace study, people compensated for interruptions by working faster, but reported greater stress, frustration, time pressure, and effort. Speed, it turns out, can conceal a fairly miserable process. (UC Irvine ICS)

Creative work contains another kind of resistance. It appears when the obvious headline has been discarded and the better one has not arrived, and it's the awkward sketch, the paragraph that is technically correct but it has no voice. It is the period in which the creator does not yet know what the work should become.

That uncertainty is not an administrative defect. It is often the work.

A writer develops judgment by hearing the difference between two nearly identical sentences, and a photographer learns by noticing what light does five minutes later. An editor moves a scene and discovers that the meaning of the next one has changed.

The artifact is being made, but so is the person making it.

No one has ever found their artistic voice while entering metadata, but

technology and those who build it rarely distinguishes between the kind of friction that is helpful to the health and happiness of creators. What typically happens is that a difficult step appears in a workflow, so the obvious objective is to eliminate it.

The result is that we retain the approval chains, tool switching, duplicated communication, and reporting rituals while removing more of the exploration, revision, and contact with the medium. Generative AI makes the imbalance especially visible because it introduces plausible answers unusually early.

Ten headlines appear before the writer has decided what is worth saying. Polished images arrive before the designer has established a visual direction. A complete first draft appears while the creator is still forming an opinion.

The output may be perfectly competent, but isn't that is part of the problem. Competent work can end the search before distinctive work has had a chance to begin.

Research suggests that this is not merely an aesthetic suspicion. In one experiment, access to AI-generated ideas helped participants produce stories that evaluators considered more creative, better written, and more enjoyable, with especially strong benefits for less-creative writers. But the AI-assisted stories also became more similar to one another. The individual product improved while the collective field narrowed. (Science)

In a separate visual-ideation study, participants using an AI image generator produced fewer ideas with less variety and lower originality than those in the baseline condition. The researchers observed that generated images could become new points of fixation: the tool supplied possibilities, but those possibilities also anchored the search. (arXiv)

This does not make AI the enemy of creativity. It makes AI one more tool that needs a job description.

The better path begins with something less glamorous than an AI strategy: examining how the work is actually done. Take a real project from request to delivery and break it into its component tasks. Not the elegant process shown in the onboarding deck. The real one, including the spreadsheet someone created during an emergency in 2021 and the approval step everyone follows because nobody remembers who introduced it.

For each task, ask three questions.

1. Does it require judgment, taste, empathy, context, interpretation, or accountability

2. Is it repetitive, rules-based, administrative, or largely a matter of moving information between systems?

3. And would removing it give creators more attention, or remove an activity through which they learn, explore, or take satisfaction in the work?

The answers suggest three different actions. Some tasks should simply disappear. A useless process does not become valuable because it has been automated.

Others are ideal candidates for technology: file routing, transcription, repetitive resizing, version comparison, asset tagging, status reporting, scheduling, format conversion, and the transfer of information between systems. These are the places where automation can return meaningful time without making creative decisions on someone’s behalf.

The most useful improvements may require teams to work with developers, automation specialists, or technically inclined colleagues who understand the actual workflow. The answer may be a modest integration, an internal script, a centralized approval record, or a system that generates routine updates automatically.

None of these will cause a sensation at a technology conference, but hey may give your designer Tuesday afternoon back.

The remaining tasks should be protected precisely because they require human attention. Problem-framing, early ideation, interviewing, writing, visual exploration, performance, revision, and final judgment should not be automated merely because automation is available.

When creators enjoy a part of the work, learn through it, or use it to develop their relationship with the medium, removing it deserves particular scrutiny. AI can still participate. It can transcribe an interview, organize research, expose clichés, generate counterarguments, compare versions, visualize a direction, or stress-test an idea that already contains human intent. It can help people travel farther without choosing the destination for them.

Leaders have to defend this distinction.

A company cannot claim to value original work while rewarding only speed, volume, responsiveness, and the number of assets pushed through the system each week. The workflow will eventually reflect the incentive, no matter how many times the word authentic appears in the strategy deck.

Protecting the creative process means limiting unnecessary approvals, creating blocks of uninterrupted time, involving creators in decisions about automation, and refusing to treat every unautomated human activity as an inefficiency waiting to be corrected. It also means recognizing that the health of the people doing the work is not separate from the quality of what they produce.

The goal is not a frictionless creative process, that would remove much of the process. The goal is to make the boring work nearly invisible and the meaningful work more possible—to spend human attention where human attention changes the result.

Before automating a task, organizations should ask more than whether a machine can do it faster. They should ask what people will have more room to do afterward.

The creative process deserves better than a future in which software makes the work and humans update the dashboard.