AI can generate a polished image in seconds. Getting that image to match your brand, work across five different formats, and survive a creative director's feedback is another story.
Adobe wants its AI agents to handle more of that work.
Ahead of Adobe MAX 2026, which begins November 10, the company is making the case that creative AI needs to move beyond generating content toward helping people finish it. As AI models improve, Adobe is betting that the ability to direct, refine, and evaluate their output will become increasingly valuable.
Its answer is the Adobe Creative Agent, the technology behind conversational assistants designed to coordinate tasks across its creative applications.
Rather than requiring creators to choose every tool and perform each step manually, Adobe wants them to describe their desired outcome, delegate parts of the execution, and adjust the work as it develops.
That could change how people use creative software. It also gives Adobe a difficult challenge: proving that automation can handle increasingly complex production work without compromising the control professionals need.
Adobe thinks creative direction is becoming the next AI skill
For years, prompt engineering has been treated as an essential skill for working with generative AI. More detailed instructions could produce better results, although users still had to manage much of the work themselves.
Adobe believes AI agents are changing that relationship.
In written responses provided ahead of MAX, the company described creativity as an iterative process involving experimentation, feedback, and judgment. A creator might begin with a specific idea, discover something unexpected in an AI-generated result, and change direction several times before arriving at the finished product.
Adobe's position is that agents can take on more of the execution while people concentrate on deciding what works.
That approach is already taking shape in Firefly AI Assistant, which entered public beta in April 2026.
Powered by Adobe's Creative Agent, the assistant lets users describe what they want to accomplish in natural language. It can then coordinate supported operations using capabilities from applications including Photoshop, Illustrator, Lightroom, and Premiere.
Consider a marketer preparing a product campaign. The job might involve editing product photography, producing social media variations, adjusting backgrounds, and preparing assets for different channels.
Instead of manually navigating every application and operation, the marketer could describe the desired outcome and use the assistant to carry out supported steps. They would still need to review the results, provide feedback, and make the final creative decisions.
Adobe has also introduced more specialized assistants within individual applications.
Photoshop's AI Assistant, for example, offers an early-beta experience for making and refining image edits through natural-language instructions. Illustrator has a beta assistant focused on production tasks such as organizing layers, updating artwork, and preparing files.
These experiences have different capabilities and levels of availability. Adobe's broader vision is to connect them into a more comprehensive creative workflow.
The potential benefit is fewer repetitive steps between an idea and a finished asset. How reliably these assistants deliver that benefit across complex professional projects remains an open question.
Adobe wants to manage the growing number of AI models
The creative AI market has another problem: There are now so many models that choosing one can become a separate task.
Some models excel at image generation, while others specialize in video, audio, or particular visual styles. Keeping track of their strengths, limitations, and costs adds another layer of complexity.
Adobe's strategy is to bring multiple models into Firefly and give creators a choice between selecting them manually and letting the assistant handle that decision.
The platform includes Adobe's Firefly models alongside offerings from partners such as Google, OpenAI, Kling, and ElevenLabs.
According to Adobe's Firefly AI Assistant documentation, the assistant can automatically select between Adobe and partner models based on the user's request unless a specific model is chosen.
That convenience comes with trade-offs.
Adobe considers content generated with its Firefly models commercially safe to use, subject to its applicable terms. When users choose or receive outputs from partner models, Adobe says they remain responsible for determining whether those models are appropriate for their projects.
Cost also matters. Firefly AI Assistant remains in beta, with access and usage limits depending on the user's account and subscription. Certain free accounts receive limited access, while most AI-powered creative actions consume generative credits for paid users. Credit usage varies by operation, and introductory rates may change.
In other words, delegating model selection to an agent doesn't eliminate the need to understand what the agent used, what the result costs, or whether the output is suitable for commercial work.
Adobe's bigger opportunity is connecting to competing AI assistants
Perhaps the more consequential part of Adobe's strategy involves where creative work begins.
A project might start in ChatGPT, Claude, or a Slack conversation rather than inside Photoshop. Adobe wants its tools to remain available wherever that work happens.
Through Adobe connectors, users can access supported creative capabilities from ChatGPT, Claude, Gemini, and Slack without leaving those environments. The integrations offer different features depending on the platform, account, and available Adobe services.
Open standards such as the Model Context Protocol (MCP), which allows AI assistants to connect with external tools, help make this type of integration possible.
The business implication is significant.
If creators increasingly use general-purpose AI assistants to plan and coordinate projects, software companies may have less control over where customers begin their work. Adobe's connectors offer a way to keep its professional tools involved even when another company's assistant becomes the main interface.
That positions Adobe to compete on the quality and depth of its creative capabilities rather than relying exclusively on customers opening its applications first.
But Adobe isn't the only company pursuing more integrated creative AI.
Figma's AI design agent can perform multi-step tasks directly on its design canvas while using existing design-system components, variables, and styles.
Canva AI 2.0 similarly brings conversational creation, editing, and brand-aware workflows into its design platform. Canva says its system produces layered, editable outputs that users can continue refining.
These approaches reflect a broader competition over how AI-generated work becomes a finished product.
Adobe has an extensive collection of professional editing tools and decades of experience developing creative workflows. Figma has a strong foundation in collaborative interface design and design systems, while Canva has emphasized accessible, integrated creation.
Each company is trying to make its existing strengths more useful through AI agents.
Adobe's challenge is demonstrating that the breadth of its software translates into better results, rather than introducing additional complexity behind a conversational interface.
The real test comes after the first draft
Adobe's argument rests on something familiar to professional creatives: A convincing first draft can still be a long way from a finished deliverable.
An image might look impressive while containing inaccurate details, inconsistent branding, or elements that don't work in the intended format. A video might need changes to pacing, composition, or individual scenes.
These adjustments are where creative judgment and precise editing tools become especially important.
Adobe believes its advantage lies in combining AI generation with the capabilities creators already use to make those corrections.
For beginners, that could mean producing work that previously required learning several complex applications. For professionals, it could mean delegating repetitive production tasks while keeping their attention on higher-value creative decisions.
But a useful creative agent needs to do more than execute instructions.
It needs to preserve editable assets, respond predictably to feedback, respect brand requirements, and let users correct individual decisions without rebuilding everything from scratch.
Those capabilities also need to work consistently enough to justify their cost.
Adobe's documentation describes ways for users to redirect the assistant, refine outputs, and take over work themselves. Whether that translates into dependable performance across demanding professional projects will be an important measure of the company's progress.
At Adobe MAX 2026, taking place November 10–12 in Miami Beach and online, the company has another opportunity to show how far its Creative Agent has developed.
The event's theme, Dream Big. Create Bigger, reflects Adobe's ambition to make more complex creative work accessible while giving experienced professionals additional room to pursue ambitious projects.
The broader shift Adobe is betting on is already visible across the industry: AI tools are moving from generating individual outputs toward helping users manage entire workflows.
Adobe wants creative direction to become the organizing principle of that work, with agents carrying out more of the execution. The test is whether creators can hand off the tedious parts and still get exactly the results they intended.
