Adobe is hosting its Max conference in Miami Beach today, where it has announced several new AI features alongside an expanded partnership with Google. Among the biggest announcements is the introduction of always-on agents for its Firefly AI Assistant, which are expected to arrive in beta in early 2027.

The new feature will allow users to select pre-built agents designed for specific creative workflows, customize their tasks, set goals, and review their plans before giving them permission to begin. Once activated, the agents can continue working in the background even when users step away. They can also determine when to repeat tasks, learn individual preferences through feedback, and request decisions or approval whenever necessary.

Adobe provided the example of a multi-week video project, where an agent could automatically organize incoming footage, remove duplicate clips, and identify potential problems such as camera shake or audio issues for the user to review. A central dashboard within Firefly will allow users to monitor progress and approve the work being carried out.

The Firefly AI Assistant is also receiving a new Plan Mode, which lets users review, edit, and approve an agent’s proposed plan before execution. Alternatively, they can allow it to proceed independently. Adobe is also introducing a Model Selector with an Auto-Select option. Users will be able to define formats, colours, brand guidelines, characters, and products once and apply those settings across their projects.

Alongside the AI announcements, Adobe has revealed an expanded partnership with Google. Acrobat will soon become available within Google Workspace through Google Drive and Gmail, appearing as a suggested application with an “Edit with” option in file previews. Adobe will also integrate Google’s AI models into the technology powering Acrobat’s AI features, expand its connectors within Gemini and Gemini Enterprise to include Acrobat and additional creative tools, and bring its creative applications to AI Mode in Google Search.

Adobe customers can already access Google’s AI models, including Gemini Omni, Veo, and Nano Banana. Both companies say Google’s latest image and video models will continue to be integrated into Firefly, Photoshop, Premiere, and Adobe Express.

Additionally, Adobe and Google DeepMind are exploring a customized Google image model for Photoshop that could enable near-real-time editing capabilities. However, neither company has provided a timeline for when the new Google integrations will become available.