Adobe took the stage at its Max conference in Miami Beach today to reveal a slate of new AI capabilities, headlined by a significant expansion of its collaboration with Google. The standout announcement is a new category of tools Adobe is calling always-on agents, which are scheduled to arrive in beta for the Firefly AI Assistant in early 2027.
These agents are designed around specific creative workflows. Users select a pre-built agent, tailor its behavior to their needs, define an objective, and approve its proposed plan before anything begins. From there, the agent operates autonomously in the background — even when you're away from your desk — determining when tasks should be repeated, adapting to your preferences based on feedback, and pausing to check in whenever it requires a decision or sign-off.
To illustrate the concept, Adobe described a weeks-long video production in which the agent could sort incoming footage as it arrives, eliminate duplicate clips, and surface problems like shaky camera work or flawed audio for human review. Progress tracking and approvals are handled through a centralized dashboard within Firefly.
The Firefly AI Assistant is gaining several other upgrades as well. A new Plan Mode gives users the choice to review, modify, and approve an agent's strategy before execution — or simply let it proceed unattended. A Model Selector with an Auto-Select option is also on the way, and users can now define formats, colors, brand guidelines, characters, and products a single time and have those settings carry over across every project.
On the partnership front, Acrobat is set to integrate directly into Google Workspace via Google Drive and Gmail, where it will appear as a suggested application offering an "Edit with" option in file previews. Adobe also plans to incorporate Google's AI models into Acrobat's feature set, broaden its connectors for Gemini and Gemini Enterprise to include Acrobat and more creative tools, and bring its creative suite to AI Mode in Google Search.
Adobe users can already tap into Google models including Gemini Omni, Veo, and Nano Banana, and both companies confirmed that future Google image and video models will continue rolling out to Firefly, Photoshop, Premiere, and Adobe Express. On top of that, Adobe is working with Google DeepMind to explore a custom-built image model tailored for Photoshop — one that could potentially deliver near-real-time editing. No launch dates have been shared for any of these Google integrations.
Source 1 | Source 2
