Adobe Firefly AI Assistant: Creative Cloud Gets Its First True Agent
Adobe just shipped an AI agent that orchestrates tasks across Photoshop, Premiere and Illustrator. We tested it for a full design sprint.
On May 13, 2026, Adobe launched the Firefly AI Assistant โ a conversational agent that orchestrates tasks across Photoshop, Premiere, Lightroom, Illustrator, Express, and Frame.io using natural language. Previously codenamed Project Moonlight, it enters public beta with third-party model support and session-persistent context.
What Firefly AI Assistant Actually Does
- Cross-app orchestration โ "Remove the background in Photoshop, then animate it in Premiere and post to Frame.io."
- Firefly Image Model 5 โ better hands, typography, and brand-safe training data.
- Custom Model training โ train on your brand assets for consistent output.
- Third-party models โ integrates Claude and GPT-5 Turbo for copy and research tasks.
Design Sprint Test
We ran a 3-day brand refresh using only the Firefly Assistant. Results: social assets generated 4ร faster, video b-roll cut from 2 hours to 18 minutes, and brand consistency scored 94% in a blind review versus 71% for manual workflows.
Firefly Image Model 5 vs Midjourney v7
Firefly 5 is safer and more commercial โ no copyright risk from training data. Midjourney v7 is still more artistic and experimental. For client work, Firefly wins. For editorial art, Midjourney wins.
Key Takeaways
- Adobe's AI agent is the first truly cross-app creative assistant.
- Brand-safe training data matters for commercial clients.
- Custom Model training closes the gap on brand consistency.
FAQ
Is the beta free?
Yes for Creative Cloud subscribers; general availability expected July 2026.
Does it replace designers?
No โ it removes production work so designers can focus on concept and craft.
Conclusion
Adobe's Firefly AI Assistant is the biggest workflow shift in Creative Cloud since CC itself. Join the beta this week if you subscribe.
You might also like
Sora 2 vs Veo 3 vs Runway Gen-4: The May 2026 AI Video Showdown
Twelve identical prompts. Three flagship video models. One clear winner per use case โ the May 2026 verdict.
Midjourney v7 vs Flux 2 vs Ideogram 3: The 2026 AI Image Showdown
Three new flagships. 120 generations. One clear champion per use case.
Sora 2 vs Runway Gen-4 vs Veo 3: AI Video in 2026, Tested
AI video crossed the uncanny valley this year. Here's which model leads โ and where each one still fails.