Adobe is doubling down on its vision to lead the integration of artificial intelligence (AI) across creative workflows. This time with a strong focus on both proprietary and 3rd party models, according to Alexandru Costin, Vice President of Generative AI and Sensei at Adobe.
AI strategy
Adobe’s current strategy, revolves around three pillars: generative models, agentic interfaces, and deep tools integration. The company is committed to bringing together its own “commercially safe” first-party models, a variety of leading partner models, and the ability for customers to create and deploy custom models.
“What our strategy is centered around, is to become the company that powers creativity with AI. And we want to do that by bringing together all the best models,” said Costin, Vice President, Generative AI and Sensei, Adobe. “We want to give them the best of both worlds and bringing some of our Titan chatgpt also very, very exciting.”
Model integration
While integrating prominent third-party models-such as those from Google, OpenAI, and others-Adobe maintains a clear commitment to its proprietary technology. Costin highlighted the importance of offering enterprise customers “commercially safe models” and explained that Adobe’s own models are designed to fill specific gaps not addressed by partners. “There is still a huge market for commercially safe models, and we think we have the opportunity by integrating third party models, to focus our first party models on areas that are not covered by third party,” said Costin.
A notable development is the release of Firefly Image Model 5, with enhanced rendering capabilities and editing options. Costin remarked on its “amazing editing capabilities at 2.4 mega pixels,” and the model is now available through firefly.com alongside new music and text-to-speech generation tools.
Partner roles
The company’s approach to third-party model selection involves strict scientific benchmarking. “We evaluate the quality of their models using our very rigorous scientific evaluation mechanisms that looks at E low ranking, technical quality, professional and professional utility,” said Costin. Adobe leverages this process to identify complementary strengths among partner models.
On the partnership with Google, Adobe perceives unique advantages: “Google, their models…have amazing video quality. So by bringing them together, this for generation in world, generation, and also, as we announced, making them available to foundry partners as another base model, we can deep tune for foundry.”
Agentic vision
Agentic AI-Adobe’s term for automated, conversational, and interactive creative tools-is already manifesting across workflows. Costin described Adobe’s intent to combine both conversational and manual interaction, especially in flagship applications like Photoshop and Illustrator. “We want to make agentic a hybrid agentic experience, where you can either converse with the app, using words and but also you can switch back and forth into full control,” said Costin.
He noted that enhanced agentic experiences will first appear in web-based interfaces before transitioning to desktop applications: “Our approach is to learn and iterate faster in the web interface. But the vision is always to bring it back to where customers needs it, but we want to take the time to mature it in the web experience, because it’s much faster to code there, and then we’ll bring it back to the desktop.”
Business model
Costin emphasised Adobe’s multi-track monetisation strategy, which includes both Creative Cloud subscriptions and usage-based credit models. “The monetization strategy will be retention and increase conversion actual seats that bundle generative credits, like Creative Cloud Pro or the Firefly SKUs,” said Costin. He added that high-volume enterprise users may pay for additional credits and API calls, with offerings tailored along the spectrum from consumers to large enterprises.
Workflow expansion
Adobe envisions its modular AI platform and APIs extending beyond Creative Cloud into enterprise and marketing applications. “There is an acceleration and compounding these modules and APIs together,” said Costin, referencing the platform’s capability to connect Adobe and third-party functionality into interoperable workflows. “This is the platform oriented vision that I see us embracing, and this is how we win.”
Technical approach
Interoperability is enabled through a unified API gateway allowing internal and external models to be accessed consistently. “We’ve built some type of a platform that enables our applications to connect through a single API gateway, and then we have a deeper connection to everybody else,” said Costin. This modular approach supports both existing creative professionals and emerging creator groups.
The relation between cloud-based and on-device AI also remains a strategic focus. While many advancements begin in the cloud, Adobe is preparing for a cycle where mature models are optimised for performance directly on local devices. “We think it’s maybe a year or two away until these models were achieved maturity in the cloud, and then they can be distilled to run on device,” said Costin.
User continuum
Costin noted that future generations of creators may engage primarily through conversational interfaces or automated workflows, potentially bypassing traditional UI experiences altogether.
“We want to empower the current customers to stay differentiated through the knowledge, and if they did school. We want them to be advantaged by all of that investment they made in their careers, and have agentic help them in that pursuit, also for new types of creators or maybe consumers that don’t have the patience to do that. We want to enable them to with models that might just do it for them,” said Costin.