{“title”: “The End of Generic AI: How Brands Are Going Custom”, “Text1”: “
Generic AI tools are losing ground to bespoke, brand-trained models, and the shift is happening faster than most marketing teams realize. According to MIT Tech Review, Adobe’s Firefly Foundry now starts with a commercially safe base model and trains further on a company’s own IP, producing content that actually reflects a brand’s vision instead of generic AI slop. The piece details Adobe’s partnerships with film studios like Wonder Studios, Promise.ai, and B5 Studios, plus the big three talent agencies (CAA, UTA, WME), all built around keeping human creatives at the center of AI scaling.
This is significant because it signals where enterprise AI is heading: away from one-size-fits-all foundation models and toward custom-trained systems that encode a company’s voice, characters, and visual language. Add Adobe’s new strategic partnership with NVIDIA for enterprise-grade compute, and you have the rough shape of the next two years in creative AI.
The agentic web is rewriting brand visibility
The second trend in the MIT Tech Review piece is bigger than custom models. AI is reshaping how customers discover brands in the first place.
The numbers Adobe Digital Insights shared are striking:
- AI-powered shopping is up 4,700%
- Agentic web traffic is up 7,851% year over year
- Most businesses have significant gaps in AI-led brand visibility
What stands out here is the inversion of the discovery model. For two decades, brands optimized for human eyeballs through Google. Now they’re optimizing for AI agents that decide what humans see. If your content is invisible to the agent, you’re invisible to the customer.
Major League Baseball is already adapting. The league uses Adobe LLM Optimizer to monitor how its content surfaces across AI interfaces and adjusts in real time. Adobe’s recent acquisition of Semrush pushes the same playbook further into search and visibility analytics.
What’s actually changing, and why now
Two forces are colliding at once. On the production side, brands need scaled, on-brand content because AI agents are consuming and surfacing exponentially more of it. On the discovery side, those same agents are the new gatekeepers between brands and buyers. Generic AI tools can’t solve either problem at the scale enterprise creative teams need.
MIT Tech Review frames it cleanly: generic AI gives teams a starting point, but a model trained on a brand’s own IP gets them to the finish line.
Practical takeaways for AI practitioners and brand teams
The article closes with a piece of advice worth repeating: audit before you automate.
- Map your content supply chain first. Who creates, who approves, where assets live, where things break.
- Identify duplicated processes and unclear ownership before plugging AI into them.
- Then layer custom-trained models on top of a clean process, not a broken one.
- Start tracking AI agent visibility the same way you track SEO. If you’re not measuring how LLMs cite or surface your brand, you’re flying blind.
AI applied to a broken process just breaks it faster.
The next 12 to 24 months
Expect three things to accelerate. First, more enterprise platforms (not just Adobe) will offer IP-trained foundation models as the default, not a premium tier. Second, “LLM optimization” becomes a budget line item next to SEO and paid social, with dedicated tooling and dedicated headcount. Third, the brands that move first on both fronts (bespoke models plus agentic visibility) will compound their lead, because AI agents reward consistency and reinforce whatever they already surface.
The teams treating AI as a faster Photoshop are going to lose ground to teams treating it as a new distribution layer. Full details at the original MIT Tech Review piece.
“}