Where Should AI End in Creative Work? The Boundary Should Stay Revisable
September 08, 2026 by Gerardo I. Ornelas
Editorial Series: Human–AI Interaction & The Trust Stack
Target Question: How can creative professionals use AI without losing authorship or agency?
Direct Answer: Creative professionals can preserve agency by deciding which tasks, responsibilities and authorship claims they will delegate to AI—and by keeping those boundaries revisable. A five-year study of 17 digital painters found that creators’ choices changed as tools and workplace pressures changed. Good creative AI should make those boundaries visible, editable and attributable rather than treating adoption as a one-time yes-or-no decision.
The hardest question in creative AI is not whether someone used it.
It is where the person decided the machine should stop.
A five-year qualitative study, submitted September 8, 2026, followed 17 Chinese digital painters through annual interviews from 2021 to 2025. The researchers observed no single march from resistance to adoption. Creators repeatedly negotiated which tasks, responsibilities, values and claims would remain human.
They called this process “longitudinal agency partitioning.”
That phrase is more useful than the familiar human-versus-AI debate because real creative work rarely fits a binary label.
Creative Agency Is Distributed Across a Workflow
A finished image can contain dozens of decisions: reference gathering, sketching, composition, lighting, background, facial expression, color, refinement, client review and final approval.
In the study, some painters eventually delegated bounded tasks such as references, backgrounds, rough sketches or client-facing drafts. Others preserved “human-only zones” around the face, the initial sketch, the final pass or a signature style.
The boundary was not identical for every creator. It did not remain fixed for the same creator over five years.
That is the central insight: agency is not a property of the final file. It is an arrangement across the process.
Adoption May Be Pragmatic Without Feeling Voluntary
Peer norms, deadlines, copyright and precarity shaped what felt delegable. Efficiency can expand capability while narrowing choice. If refusal is costly, adoption is not purely individual; tools should measure more than speed.
The Boundary Should Be Explicit
Most creative AI products offer controls over the output: prompt, style, variation and strength.
They offer far less control over authorship boundaries.
A better interface would let a creator state:
- AI may generate reference material, but not final pixels.
- AI may propose backgrounds, but I approve every selected element.
- My sketch and signature style are protected zones.
- Client deliverables must show which stages used AI.
- This boundary applies to this project, not every future project.
These are not simply settings. They are declarations of creative responsibility.
The system should preserve them in the project history and allow the creator to revise them as the work evolves.
Revision Is a Feature
Creators responded to changing capability, economics and community norms. Revising a boundary can be reflection, not inconsistency. Software should show the current arrangement, earlier versions and what changed.
Provenance Should Describe Contribution, Not Deliver a Verdict
Creative provenance is useful when it helps someone understand who contributed what and when.
It becomes harmful when compressed into a badge that declares an artifact human-made or AI-made.
The painters’ workflows show why. A person may direct the concept, make the sketch, use AI for references, paint the central subject, approve the background and perform the final pass. One binary label cannot represent that distribution.
The goal is not a moral score. It is a legible record of contribution and responsibility.
This is where the Human–AI Interaction series meets The Trust Stack: The Provenance Spectrum: provenance can help preserve the path, but the human question is whether the creator retained meaningful choice over that path—a core principle of human agency in the future of work and verifiable human intent.
Original Synthesis: The Boundary Is Part of the Work
We tend to treat the boundary between person and machine as evidence collected after creation.
I think it is part of the creative act itself.
Choosing what to delegate, what to protect, what to approve and what to claim is now one of the decisions that gives a work its shape.
The best creative AI will not make that boundary disappear. It will help a person draw it, inspect it and redraw it.
What Builders Should Design
- Support stage-level delegation: Allow creators to assign AI assistance to discrete subtasks rather than all-or-nothing generation.
- Human-only zones: Provide protected phases (e.g., initial sketch, character faces, final polish) where generative mutations cannot touch pixels.
- Revisable controls: Make workflow boundaries editable dynamically as project needs shift.
- Contribution history: Log who and what shaped each layer, preserving provenance as a structured timeline.
- Team norms & policy configuration: Enable studios and teams to establish clear, shared delegation baselines.
Limits of the Evidence
This qualitative study covers 17 painters in one context; interviews do not measure productivity. Its depth supports a hypothesis to test: explicit, revisable boundaries should increase felt control.
A Better Question Than “Did You Use AI?”
The binary question is easy for platforms and poor for people.
A more revealing set is:
- What did the person intend?
- What did the AI contribute?
- Which stages remained human?
- Who approved the final result?
- Could the creator change the boundary?
Human creative agency will not survive because we preserve a clean separation from machines. It will survive when people retain the power to decide where that separation belongs.
Primary Sources
- Meng et al. (Sep. 8, 2026): Longitudinal Agency Partitioning in Human-AI Creative Work
- Related Analysis: The Provenance Spectrum: Why a Label Cannot Protect Human Creativity by Itself
- Related Analysis: Labor Day in the Age of AI: The Future of Work Is a Human Agency Problem
- Related Analysis: Verifiably Human — Part II: The Death of Ambient Authority
Frequently Asked Questions
What is creative agency in AI workflows?
Creative agency is the ability to set creative goals, decide which specific tasks and responsibilities to delegate to AI, and retain final authorship and judgment across the entire workflow.
Does using AI erase authorship?
Not necessarily. Contribution, control, stage-level human involvement, and context matter far more than binary labels. Authorship is preserved when the creator directs intent and approves consequential choices.
What is a human-only zone?
A human-only zone is a specific stage, asset, or decision (such as the initial sketch, facial features, or final polish pass) that is explicitly reserved for human execution and protected from automated AI transformation.
© Gerardo I. Ornelas
Systems architect, founder, and advisor for governed AI and trusted visibility.