Adobe’s newest artificial-intelligence announcements move the creative question beyond generating a single image or editing one clip. The company is building systems that can plan and execute sequences of work across Firefly and Creative Cloud, including an expanded Creative Agent, repeatable Firefly Graph workflows and future connections to third-party agents.
The shift is from tool use to delegation. An artist can describe an outcome, review a proposed plan and allow software to coordinate several operations. That may reduce repetitive production, but it also changes where creative control, rights checks and responsibility need to sit.
What is available—and what is still a preview
Adobe says the Firefly AI Assistant beta is gaining a Plan Mode that lets users review, edit and approve a sequence before it runs. A Model Selector can specify an AI model or allow automatic selection. Firefly Graph is designed for custom, repeatable workflows rather than isolated prompts.
Other capabilities are not yet generally available. Local Model Context Protocol integrations, previewed for a private beta in early 2027, are intended to let external agents work with native files in applications including Photoshop, Illustrator, Premiere Pro and After Effects. Adobe also previewed “always-on” agents for recurring background tasks but did not announce a public launch date.
That distinction matters. Artists evaluating a production pipeline should separate a released feature from a beta, a private beta and a concept preview. A promised integration is not yet a dependable delivery system.
The approval step becomes a creative boundary
Plan Mode may prove more consequential than another generative effect. When an agent proposes several steps, the artist can inspect the sequence before execution. That moment is an opportunity to decide which choices are administrative, which are aesthetic and which require human judgment.
A plan should not be approved as one opaque block. Creators can ask: Which source files will be accessed? Which model will be used? Will material leave a local machine? What changes will be written into the native file? Can the work be reversed? Does any step create new content that needs disclosure or rights review?
Editable layers, vectors, bins and sequences help preserve control after an automated pass, but editability is not the same as authorship. A creator still needs to know why a change was made, whether it reflects the brief and whether the inputs were authorised.
A practical protocol for artists
Before putting an agent inside a client workflow, create a short operating record. Note the approved source assets, the selected model, the plan, the human checkpoints and the final manual changes. Keep the record with the project files.
For commissioned work, add explicit terms covering AI-assisted processing, third-party services, confidential assets and client approval. If the agent can connect to documents, messaging or storage, grant only the access required for the task and remove it when the project ends.
Test repeatable workflows on non-sensitive material first. Compare output across several runs. Check whether automatic model selection produces consistent results and whether brand or style preferences are applied too broadly. An efficient error repeated at scale is still an error.
Delegation without disappearing
Agentic tools can be valuable for versioning, asset preparation, formatting and other production-heavy work. They may also make complex applications more accessible to creators who cannot master every interface. The risk is not automation by itself; it is automation whose decisions, permissions and limits are invisible.
The strongest creative workflow will keep the artist present at the points where meaning changes. Software can coordinate steps. The creator remains responsible for deciding which steps should exist, what evidence should be kept and when a machine’s plan should be refused.
ATIM approaches artificial intelligence with curiosity, independence, and balance. We explore its extraordinary potential to empower artists while examining the questions it raises around authorship, copyright, consent, creative identity, transparency, and human expression.
We are neither advocates for AI nor opponents of it. Our focus is how the technology is used—and what that means for artists.
Above all, our commitment is to artists. We look at every development through the lens of what can support, protect, and empower the creative community—identifying opportunities, examining potential risks, and providing artists with the information they need to make informed decisions about their work, their rights, and their future.