Editing a photograph can take seconds in your imagination and much longer at the keyboard. Move a light source, remove an object, resize the design and prepare a version for a different screen, all without losing the original layers. Adobe's latest Adobe Creative Agent announcement is designed to shrink that distance between intention and execution, as the company expands conversational assistants across Creative Cloud and Firefly.
At its MAX event on 10 November 2026, Adobe described new assistants, editing controls and preview integrations that would allow AI agents to work with original project files. The distinction is vital: Adobe is announcing an expanding mixture of shipped tools, betas and previews, not claiming that every Photoshop user can already hand over an entire production job. Its official MAX announcement describes the direction clearly, while the actual availability depends on the application and feature.
Adobe Creative Agent starts with less clicking
Many creative jobs contain an unglamorous second job: preparing assets. A designer might know exactly what a poster needs yet spend time finding layers, adjusting shadows or rebuilding the layout for a social platform. A video editor may need the same scene in several sizes, with colour correction and captions checked for each export. Those tasks require skill, but not all of that skill belongs in repeated clicking.
The expanded Creative Agent is Adobe's attempt to let someone describe a multi-step goal and supervise the result. Assistants are being extended across applications including Illustrator, Premiere, InDesign, After Effects and Frame.io. The company is also previewing connections that let preferred third-party agents direct edits inside Photoshop, Illustrator, Premiere and After Effects, rather than simply generating a disconnected picture or video file.
That last detail matters more than a dazzling demonstration. Professionals need to reopen a project, adjust one element and deliver a revision without rebuilding the whole thing. Adobe says its local integrations are intended to preserve editable layers, sequences and compositions. But the integrations are previews. Teams should test them before promising a client that an agent can reliably handle a production workflow.
Photoshop gains a different sort of lighting assistant
Adobe's product walkthrough describes Relight, which places and adjusts new light sources with non-destructive controls. That is different from typing a prompt and accepting a completely remade scene. A photographer could experiment with where illumination falls while retaining some control over the original image. Generative Restyle offers prompted visual variations with a slider to adjust intensity, while other beta features seek to simplify changes in camera angle and composition.
There is a sensible reason Adobe keeps emphasising manual precision. An attractive AI-generated result may be useless if it alters the identity of a product, changes the lettering on a sign or moves an essential detail. In commercial photography, a small mistake can be more expensive than a dramatic visual improvement is valuable. The ability to inspect and reverse changes remains part of the job.
AI Assisted Editor offers a simplified starting point inside Photoshop, with the possibility of switching back to professional editing controls. That arrangement may help occasional users, but it also changes the expectations placed on expert editors. The computer may handle the first pass. The human still needs to see whether an edge is clean, whether the colours are consistent and whether the image actually tells the intended story.
Firefly becomes a workbench rather than a single generator
Firefly is also becoming a broader creative environment, spanning image, video, audio and design tools. Adobe describes a choice of partner models rather than insisting that every task uses one engine. New features include editing part of an image without changing the rest, guided quick actions and systems for building repeatable creative workflows.
One preview involves agents that can continue recurring background tasks, such as organising assets or preparing versions for different channels. Another lets users construct a process they can run repeatedly. For a small publishing operation, that could be more valuable than asking AI to invent a completely new design every morning. Consistent crops, file preparation and quality checks make it easier to publish at volume without overlooking basics.
There is a qualification: background agents require careful instructions, permissions and quality checks. An assistant told to prepare a campaign should not be free to alter a signed-off logo or quietly replace legally approved wording. A reliable system needs clear boundaries, previews and human authority over the final release. Speed only creates value if the output remains usable.
The creative profession is still being asked to judge
The industry's longstanding debate about whether AI can truly be creative has often centred on whether a machine can produce an impressive image. Adobe's new tools shift the question towards who decides what should be made and who is accountable when the output is wrong. Art direction becomes a conversation with software, but the software cannot understand a client's preferences simply because it can accept an instruction in plain English.
Professional designers also bring history to a project: why a brand uses particular colours, which elements must stay consistent and where a campaign is allowed to take risks. Those decisions may be written down in guidelines, yet even the best guide leaves room for judgement. The agent's value lies in translating decisions into revisions, not in declaring that judgement unnecessary.
For artists worried about being displaced, the issue is real rather than theoretical. Faster first drafts can change what clients are willing to pay for. Equally, an editor who can deliver precise, distinctive work across multiple formats may become more useful. As our earlier look at AI in the creative industries explored, the economic consequences depend on how organisations use the technology, not simply on the existence of a new button.
What to test before promising faster delivery
For anyone considering the tools, the first experiment should be small and measurable. Take a familiar project, retain a clean original, give the assistant a narrowly defined task and compare the time spent against a manual edit. Look especially closely at layers, fonts, colour, file size and consistency between exported versions. Ask whether correcting the agent's mistakes took longer than doing the work directly.
Adobe has shown a more integrated future in which agents operate inside the creative files people already use. That is an appealing prospect for anyone drowning in repetitive production work. Its credibility will depend on whether a designer can trust not only the first impressive result, but also the tenth change requested just before a deadline.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.