From: redpointai
Adobe’s AI Strategy and Product Development
Adobe’s Chief Product Officer and Chief Strategy Officer, Scott Belsky, discussed the company’s approach to AI integration in creative tooling, covering images, video, music, and more [00:00:00]. The conversation highlighted when companies should train their own models, startup opportunities in the creative tool space, the future of Adobe AI, and what still needs to be built for enterprises [00:00:13].
Adobe’s Core AI Philosophy
Adobe focuses on interfaces, models, and the underlying data [00:03:00].
- Models: The Firefly family represents Adobe’s homegrown set of generative models [00:03:10]. These models are trained exclusively on licensed content, and Adobe has a compensation program for content contributors [00:03:17]. This approach aims to provide indemnification for customers, ensuring compliance and addressing legal concerns [00:03:29]. For example, Firefly prevents generation of copyrighted characters like Spider-Man, which is a feature not a bug, due to legal implications [00:04:51].
- Interfaces: Adobe integrates these models into products like Photoshop [00:03:34]. They also make models available to third parties and large companies for complex, at-scale workflows [00:03:42].
- Data: A key focus is on enabling custom models for customers [00:04:06]. For instance, Nickelodeon could train a version of Firefly on SpongeBob SquarePants to generate content for character development and storylines without concern [00:04:12].
- Partnerships: Adobe partners for LLM capabilities, surfacing them through products like Acrobat and digital experience products for marketing analytics [00:04:28].
Deciding to Build Proprietary Models
Adobe’s litmus test for building its own models is whether it can deliver a better end-to-end experience than anyone else [00:11:33]. This includes fine-tune controls, custom model capabilities, and leveraging user data on tool usage to refine model output [00:11:51]. If an area is commoditized or another company will always invest more and have something better, Adobe partners with them [00:12:41].
Product Development and User Experience
Adobe is focused on how creative professionals can unlock new workflows [00:01:05].
- Project Neo and Firefly Integration: Adobe observed users combining Project Neo (a 3D illustration program) with Adobe Firefly to achieve extreme precision in image generation [00:01:15]. This unexpected workflow highlighted that future AI advancements are less about the “next best model” and more about the controls applied on top of the model [00:02:07].
- Expanding Exploration: AI-driven tools enable creative professionals to explore a far wider range of possibilities much more quickly, overcoming time constraints [00:06:27].
- Generative Recolor: In Illustrator, this feature allows users to apply hundreds or thousands of color palettes based on prompts to vector creations instantly, a task that previously took days [00:06:42].
- User Education and Taste: The availability of new “superpowers” through AI doesn’t mean users know how to wield them [00:07:49]. Just as photography evolved to emphasize lens choice, lighting, and photo selection, taste becomes a crucial skill that AI offloads rudimentary tasks [00:08:29]. Adobe’s goal is to enable users to flex their taste through prompt augmentation and onboarding experiences [00:09:07].
- Future UI: The next generation of AI will involve models not only generating content but also providing custom UI on demand for fine-tuning, which disappears when the task is complete [00:13:22]. This means the UI would adapt to the specific use case, rather than being static [00:13:58].
- Professional Grade Quality: The bar for professional-grade media creation is high, especially for video [00:14:17]. Adobe focuses on pragmatic, high-value use cases, such as extending a video scene by a few seconds in Premier Pro to save re-shooting a whole scene [00:15:30].
“The devil’s in the defaults.” [00:28:16] Making AI-driven tools a default part of the product interface, like Generative Fill in Photoshop, significantly increases utilization [00:28:22]. This requires a different product playbook than simply hiding new features in a panel [00:28:50].
Business Model and Pricing
Adobe uses a generative credit system [00:26:01]. Customers receive a basic amount of credits with their existing plans, with options to upgrade for more [00:26:48]. This allows Adobe to adjust credit costs for more intensive capabilities, while aiming to keep them as cheap as possible to ensure accessibility [00:27:17].
Market Trends and Opportunities
- Hyper-Personalization: The creative and marketing space is moving towards hyper-personalized digital experiences at scale, where websites and emails will be tailored to individual users [00:17:22]. This requires integrated workflows, customer data platforms, and creative stacks with brand check AI [00:17:43].
- Startup Opportunities: Significant opportunities exist in enabling small businesses to operate with the efficiency of large ones [00:19:04]. This involves tackling cumbersome processes in areas like law or government [00:19:29].
- Future of Models: Scott Belsky initially believed a few models would dominate, but now anticipates thousands of models, many niche, some commoditized, and others focused on specific use cases [00:29:07]. Adobe’s tools should allow users to choose different models for different use cases, for example, using less commercially safe models for ideation like mood boards [00:10:17].
- Impact on Creativity: While content generation will become free and ubiquitous, humans will increasingly crave scarcity, meaning, and craft in digital experiences [00:20:50]. This will reboot the role of humans in content creation [00:21:58]. Shared social experiences, like discussing entertainment, are unlikely to disappear [00:22:04].
- AI in Music: Many music startups overemphasize AI’s role and overlook that a song’s value often comes from its human story [00:22:28]. AI can lower the barrier to music creation, but enabling artists to tell human stories behind their creations is crucial for resonance [00:22:56].
- Proactive Suggestions: The next tier of AI tools will move beyond simply doing things for users to offering proactive suggestions, guiding users to explore new avenues or anticipating performance issues in different contexts [00:24:25].
- OpenAI and Frontier Models: Adobe views frontier model players like OpenAI and Pika/Sora as part of the ecosystem [00:30:18]. As long as people need interfaces to tell stories, leverage data to customize, and use tools to achieve their vision, Adobe maintains its strong position [00:30:23]. Improving models are seen as platforms that enhance Adobe’s products [00:30:33].
Scott Belsky’s Side Hustle
Scott Belsky publishes a monthly newsletter called “Implications,” which serves as an accountability mechanism and a way to connect dots from his learnings and conversations [00:34:17]. He does not use AI for writing it [00:35:00].