Three days in LA. Thousands of the world’s most creative individuals. One company betting the future of design on machines that can think. My wife, , and I watched Adobe bet its entire Creative Cloud on AI, and walked away realizing the real story wasn’t about tools at all. It was about what happens when creativity itself starts to automate.
My first thought wasn’t about design trends or creative workflows. It was about architecture: How is Adobe planning to integrate AI across their entire stack without creating a security nightmare?
I’m wondering things like: “Where does the data go, and who owns what you create?”
Adobe MAX 2025 gave us answers: some reassuring, some concerning, and some that reveal where creative work is headed whether we’re ready or not.
TL;DR:
Adobe’s Big Pivot: They’re rebuilding Creative Cloud around AI, not just adding AI features.
Agentic Assistants: Photoshop, Express, and the new Project Moonlight act like autonomous creative agents that handle entire projects.
Project Graph: A new node-based “glue layer” linking apps into modular, automated workflows — creative GitHub for designers.
Firefly 2.0: Now a full production studio with custom models, multimodal editing, and integrations with OpenAI, Google Veo, and others.
Human Value Shift: Your judgment and creative direction now matter more than technical execution.
The Big Question: Where’s the line between AI assistance and AI replacement? It’s moving faster than most creatives realize.
The Shift Nobody Saw Coming
Let me set the scene: Over ten thousand creatives packed the Los Angeles Convention Center. The energy was electric, undercut by tension.
Everyone knows AI is coming for creative work. The question hanging in the air: Is Adobe here to help us adapt, or are they building the tools that make us obsolete?
Day One Keynote
They’re not adding AI features to Creative Cloud. They’re rebuilding Creative Cloud around AI.
That’s not a subtle distinction.
What Adobe Actually Announced (And What It Means)
Forget the hype for a second. Here’s what changed:
Agentic AI Assistants Across the Suite
Adobe announced AI assistants for Express, Photoshop, and an orchestration layer called Project Moonlight.
You don’t navigate menus anymore. You describe what you need.
“Remove the background and adjust the lighting” isn’t a four-step Photoshop process. It’s a sentence.
Adobe Express AI Assistant (public beta now): You can describe branded content and it generates it using your actual brand assets. Not templates that look vaguely on-brand—your exact colors, fonts, logos, style.
Photoshop AI Assistant (private beta): Lives inside Photoshop, understands your workflow, suggests edits based on how you actually work. It’s less “AI that knows Photoshop” and more “AI that knows how you use Photoshop.”
Project Moonlight (private beta): This is the ambitious one. It’s an orchestration layer that coordinates across multiple Adobe apps. You describe a project—”I need social content for our product launch, use the photos from last week’s shoot, match our brand style, create versions for Instagram and LinkedIn”—and Moonlight handles it. Pulls assets from Creative Cloud, generates content, sizes everything correctly, even analyzes what’s been performing on your channels to inform the approach.
Project Graph is Adobe’s most quietly radical announcement—a new, node-based workflow system designed to unify Creative Cloud into one programmable ecosystem. Think of it as the “glue layer” that connects Photoshop, Premiere, After Effects, Illustrator, and even Firefly through visual nodes that handle data, assets, and automation logic.
Here’s what that means in practice: instead of jumping between apps or exporting files manually, you can create reusable, modular workflows that pass content and instructions between tools automatically. Each node represents a creative or computational step—an AI model, a filter, a render, a brand asset lookup, or even a publishing action.
For example, a video editor could build a Project Graph where:
Firefly generates new background art in your brand style
Photoshop automatically applies adjustments
Premiere imports the asset into a sequence
After Effects renders social cutdowns
Then Express posts final assets directly to a campaign folder
All of this happens as a single, visual graph you can save, share, and remix across teams—like creative GitHub for workflows.
Under the hood, Project Graph hints at something bigger: a shift toward modular creative automation, similar to Unreal Engine’s Blueprints or Blender’s Geometry Nodes. It turns Creative Cloud from a bundle of apps into an interconnected platform. And since Adobe has been quietly building APIs for every major app, Project Graph likely sits atop that API network, orchestrating assets, AI models, and human inputs in one place.
In short, Project Graph isn’t just a feature—it’s Adobe’s attempt to make the creative process itself a first-class object in their ecosystem. It’s the architectural foundation for the next decade of Creative Cloud automation.
The part nobody’s talking about: These aren’t just chatbots. They’re autonomous agents with access to your entire Creative Cloud library.
That’s powerful. That’s also a massive trust surface.
Firefly Becomes the Central Hub
Adobe redesigned Firefly from “that AI image generator” into “your entire creative production studio.”
New capabilities:
Firefly Image Model 5: Better photorealism, 4MP native resolution, handles complex compositions without mangling anatomy. Expected upgrade, nothing revolutionary.
Prompt to Edit: Point at something in an image and say “move this” or “change that.” It edits precisely without regenerating the entire image. This is actually useful for iteration.
Video Timeline Editor: Mix AI-generated footage with your own clips. Generate transitions between shots. Fill gaps in your footage with AI that matches your style.
Audio Suite: Generate soundtracks matched to video length. Create voiceovers in multiple languages. Clean up noisy audio. Generate custom sound effects.
Custom Models: Upload images of your work, and Firefly trains a model on your style. Then generates new assets that match your brand.
The strategic play: Adobe integrated AI models from Google (Veo3, Imagen3), OpenAI, Runway, Luma AI, ElevenLabs, and others. They’re not trying to win the “best model” race. They’re building the platform where all models live.
The Questions People Aren’t Asking (But Should Be)
Day Two was packed with product demos and feature showcases. True to The Secure Circuit spirit, I had a few questions!
Question 1: What Actually Happens to Your Data?
What Adobe emphasized: “Your data stays yours. AI assistants access your Creative Cloud library with permission.”
What they glossed over: There’s a massive difference between “not training public models” and “not analyzing your data.”
Custom Models literally require analyzing your uploaded work to create a style model. Firefly Boards needs to understand your brand assets to generate consistent content. Project Moonlight examines your social performance data to make recommendations.
That’s not inherently bad: it’s functionally necessary. But here’s what you need to understand:
When you give AI assistants access to your Creative Cloud library, you’re creating a trust surface that didn’t exist before. Adobe has strong incentives to handle this correctly: enterprise clients won’t tolerate data leaks. But trust isn’t verification.
Action items:
Audit your Creative Cloud privacy settings now, not later
Understand the difference between personal accounts and team/enterprise accounts
Read the terms for Custom Models specifically—they’re different from standard Creative Cloud terms
Know which AI features require data upload vs. local processing
Question 2: What Does “Commercially Safe” Actually Mean?
What Adobe emphasized: Firefly models are trained on licensed content and public domain work. Content Authenticity web app provides attribution and provenance tracking.
What they didn’t explain: “Commercially safe” is Adobe’s marketing term, and it’s brilliant positioning. But let’s decode it.
It means Adobe will indemnify you if someone claims your AI-generated content infringes copyright. That’s a legal shield: Adobe will defend you in court.
What it’s NOT: A technical guarantee that generated content won’t resemble existing work.
AI models create novel combinations of training data patterns. Even with licensed training data, there’s no way to guarantee zero resemblance to existing creative work. Adobe’s betting their legal team is more valuable than absolute technical certainty.
My read: This is Adobe’s competitive advantage over Midjourney or Stable Diffusion. Enterprises need legal protection more than they need cutting-edge quality. Adobe’s providing the thing that matters for procurement decisions.
That’s valuable. Just understand what you’re actually getting.
Questions to answer before deploying these tools:
What actions can AI assistants take autonomously?
What requires explicit approval?
How do you audit what actions were taken?
Where does liability sit when AI-generated content causes problems?
Adobe’s building powerful tools. They’re not (yet) building comprehensive answers to these governance questions.
The Enterprise Adoption Pattern Adobe’s Banking On
Here’s what became clear watching Adobe’s positioning:
Enterprises want AI capabilities. They’re terrified to deploy them.
It’s a pattern I see constantly in my work: Creative teams want these tools. Security teams won’t approve them. Legal teams have questions nobody can answer. IT wants to know how this integrates with existing governance.
The blockers aren’t technical. They’re trust-based.
Adobe’s entire MAX pitch was essentially: “We understand enterprise requirements. We’re building AI that you can actually deploy without getting fired.”
That means:
Data governance: You control where your content goes (in theory)
Commercial safety: Legal protection if AI-generated content causes problems
Audit trails: You can see what the AI accessed and when (assuming you know where to look)
IT integration: SSO, admin controls, compliance frameworks that procurement departments understand
What I Actually Learned (Beyond the Keynotes)
covered the Adobe Max event as well on her Substack.
But here’s what we discovered talking to attendees between sessions:
Most people weren’t excited about the tech. They were anxious about the shift.
The designers and illustrators we talked to had a recurring concern:
“If AI can generate assets in my style, what’s my value?”
Fair question. Uncomfortable answer: Your style isn’t your value. Your judgment is.
AI can replicate your technique. It can generate infinite variations in your visual style.
What it can’t replicate:
Knowing which concept will resonate with a specific audience
Understanding brand strategy and positioning
Making creative decisions that reflect taste, not just technical execution
Navigating client relationships and feedback
The creatives who survive this aren’t the ones with the best technical skills. They’re the ones who know how to direct AI to execute their vision.
That’s a different skill set. And Adobe MAX 2025 was essentially a 10,000-person crash course in adapting to that reality.
What You Should Actually Do With This Information
Adobe MAX wasn’t just a product announcement. It was a preview of where creative work is headed.
Here’s what matters:
If You’re a Creative Professional:
Stop competing on execution speed. Start competing on judgment.
AI will handle asset generation, variations, resizing, format changes. Your value is knowing what’s worth making and why.
That’s not a threat—it’s a filter. The creatives who thrive are the ones who understand: AI augments taste, it doesn’t replace it.
Learn to direct AI effectively. Not prompt engineering (that’s overrated). But clear communication of intent and efficient iteration.
If You’re Building or Maintaining Creative Workflows:
Start planning for AI-native workflows now.
These tools are shipping. Creative Cloud Pro subscribers get access to AI assistants and Firefly Boards today. Project Moonlight rolls out soon.
Your team will start using these tools:officially or unofficially. Better to get ahead of it with governance, training, and policies than to react after someone uploads confidential client work to a beta AI assistant.
Questions to answer before that happens:
What data can AI assistants access?
What actions can they take autonomously?
How do we audit AI usage across the team?
What’s our policy on AI-generated content for clients?
If You Run IT or Security for Creative Teams:
Adobe’s building the enterprise AI creative stack whether you approve it or not.
Your designers are already using AI tools: probably consumer versions without IT oversight. Adobe MAX showed that enterprise-grade creative AI is here.
Your choice: Control it proactively with approved tools and governance, or react to shadow IT when something inevitably goes wrong.
Adobe’s betting you’ll choose the former. They’re probably right.
The Uncomfortable Question MAX Left Unanswered
Three days, 250+ sessions, countless demos. But one question kept coming up in hallway conversations that never made it to the keynote stage:
“Where’s the line?”
At what point does AI assistance become AI replacement?
When does augmenting creativity become automating it?
When does “helping designers work faster” become “designers aren’t needed”?
Adobe’s answer: AI enhances creativity, it doesn’t replace it. Tools empower artists. Human judgment remains essential.
The uncomfortable reality: That’s true until it’s not.
Right now, AI handles repetitive execution work. Freelance designers still make creative decisions. But Project Moonlight is described as an AI that “suggests content strategy based on social performance.”
That’s not execution. That’s strategy.
The line is moving. And Adobe’s not going to tell you when it crosses from “assistant” to “replacement” because that’s not a question you ask at your own product launch.
Read ’s Adobe MAX Sneaks coverage:
Subscribe to The Secure Circuit for the technical deep-dives, security analysis, and architectural breakdowns.
Join the AI Flow Club if you want hands-on training for these workflows. TechTiff and I are building courses covering AI-assisted creative work, automation that actually works, and security-first AI integration.
Also Check Out A Relevant Previous Post:
