Main content

Adobe Summit 2026 - NBCUniversal outlines its approach to rollout of AI agents and assistants as Adobe rolls out new tools

By Madeline Bennett April 22, 2026

Your browser doesn’t support HTML5 audio

Dyslexia mode
Excerpt:
New AI tools create marketing campaigns in minutes and turn flat images into 3D objects


Artificial Intelligence (AI) is dominating this year’s Adobe Summit, and in Las Vegas this week, the firm has been showing how its new agentic AI and AI assistant tools help businesses massively speed up work tasks and capitalize on trends.

One of the most eye-catching demos featured Adobe GenStudio and the new Adobe Brand Intelligence. The company showed how it’s possible to go from brief to campaign in under 10 minutes, something that would have taken marketing teams weeks without AI assistance.

The first step in the brief-to-campaign process is to enter a prompt into GenStudio – in this case, to take a strategy brief with new audiences, use it to develop a creative brief, pull in the right assets and package it up for the creative team. Once the link is added to the strategy brief in GenStudio, the tool immediately starts pulling a plan together.

Adobe Brand Intelligence offers insights such as which audiences to add, and why this creative brief would resonate with them. Once the marketer has chosen their preferred audiences, the full creative brief is ready to go. According to the Adobe marketing team, this task used to take them three weeks; now that’s down to seven minutes.

After asking the creative agent to find assets that were created previously that could be relevant to this new campaign, and generate more on-brand concepts to use, the brief is packaged up as a Firefly board and handed to the creative team. They can then pick this up directly in Firefly Boards to finish the concepting phase before moving on to design.

Image manipulation

The demo showed how designers can use Adobe technology to manipulate text in photo images, in this case changing the title of a book that was in a specific font and colour. Using Firefly, the designer was able to select the text and the technology could understand the context of the rest of the image, and blend both the text and the image together. Normally, that would need to be done in Photoshop using advanced skills to understand how to blend and know what the font is (a designer sat behind me in the audience remarked this would normally take them a day to complete such a task).

Going from concepting in Firefly to content creation in Photoshop, we were shown how a 2D image of a cup of coffee can be dropped into the marketing photo with one click, and with the help of Adobe’s new Rotate Object available in Photoshop beta, it’s turned into a 3D object. What was a flat image can be rotated, tilted and displayed in a different angle. Another Firefly feature called Harmonize uses AI to add suitable shadows, colours and tones to match the rest of the image and blend everything together.

Back in Frame.io, the asset is analyzed via Brand Intelligence, which reports back any items that don't comply with a company’s brand guidelines.

Frame.io also gives marketing and creative teams a simple way to collaborate. Once feedback is given and changes are requested and approved, the creative agent is able to digest all the comments and make the required changes to images and text, a task it completed in seconds.

NBCUniversal 

NBCUniversal (NBCU) has been using Adobe’s Firefly tools for these kinds of capabilities. At Adobe Summit, Ashish Desai, EVP, AI & Enterprise Innovation at NBCU, explained how as the business has been thinking about AI, it’s found the prospect both really exciting and really complex to navigate. He noted:

We've always been a company that leans into technology to advance our storytelling, to enhance our experiences and content, and we're also a company that never compromises on our IP, our values, and the trust and respect of our creators and partners.

The fact the Firefly model is commercially safe, with no scraped content, and only licensed IP and public domain content, is a huge differentiator for NBCU. Desai added:

It allowed us to have a safe place to first experiment with AI across creative workflows and ultimately scale across the entire enterprise.

NBCU now has over 2,000 creatives using Firefly in real production workflows, including for promos and campaign assets.

The media and entertainment firm has a goal of building an AI-ready organisation, and is approaching this through enablement programmes that include access to tools and technologies, education curriculum and focus workstreams.

Non-prescriptive approach

Firefly was the foundation for this strategy, and the business took a decision to be opinionated about the platform and the technology but not prescriptive about the way it was being used by the creative team.

NBCU provided broad access to the technology, and its creatives gave feedback on how they were applying it to their day-to-day workflows, their marketing, graphics or promotional materials, showing how AI could transform the way they were working. This included everything from styling logos by seasonality, to new artwork and assets for franchises and new content releases, to entirely new ways of concepting and storyboarding for studio set and production designs. Desai said:

With something like Firefly, you're able to much more quickly iterate on those design concepts and fail fast and succeed fast as well.

As the pressure grows on marketing teams to provide more assets, more personalization, and more customization, achieving that at scale in an enterprise like NBCU can be difficult. The Custom Model framework inside Firefly is helping the company here, for example letting NBCU’s streaming service Peacock take the features it built for the NBA, and automate and scale those for the Olympics. Desai said:

We loaded every single asset that was generated for the NBA, so this was hard work by the creatives, it had their intent there, the brand guidelines, the styles, the image intent from all the marketers and the product managers, and then we uploaded it into Firefly custom models and automated the creation of all that for the Olympics. So now everything you saw on features like Rinkside Live, associated with all of the Peacock front ends, these were elements that were created out of a foundation that didn't exist before, leveraging the general animation and image and video capabilities of Firefly, but grounded in the context of what our marketing and product teams wanted.

Coworker

The other demo that stood out was of the new Adobe CX Enterprise Coworker. According to Adobe:

Coworker will simplify the process of bringing together data and content across fragmented systems to deliver personalized experiences at scale, with AI agents that can orchestrate workflows based on defined goals.

Summit delegates got to see this in practice, with a demo using Adobe customer Ulta Beauty, where a small team of just a couple of people working together with agents were able to drive a global campaign in minutes using external social trends.

CX Enterprise Coworker gathers its intelligence from three main sources: Adobe applications, such as Real-Time CDP and Adobe Journey Optimizer; Ulta's own sources, like inventory, data, and operations; and the external world, including social signals, trend data and real-time events. Through the Adobe technology, Ulta has access to a variety of agents, and one of these is a custom social media agent, which is constantly scanning for viral trends that Coworker can capitalize on.

Viral video

The demo used the example of a beauty influencer posting a recommendation for using two Ulta products together, which goes viral; within hours, the video has four million views. The issue is, this all happens during the middle of the night in the US. Step up Coworker, which gets a signal from the social media agent and rallies the rest of the agents into action.

In seconds, it reaches into real-time customer data platforms and gathers intelligence about who has purchased or expressed interest in this product; it clusters existing audiences and suppresses conflicting ones; and it then leverages Firefly Creative Production Agent and Journey Optimizer to determine what message to give, what content, and what offer based on what has worked well in the past.

Once the plan is outlined, it simulates projected outcomes and confirms that governance conditions are met before seeking approval. All this with no human intervention so far.

Because it’s the middle of the night in the US, Coworker knows to reach out to the UK marketing group, who are able to review the plan and quickly approve it. Two campaigns launched across email, push, and paid social in minutes. Adobe noted that responding to a viral moment like that pre-agentic AI could have taken days and might have missed the moment.

Adobe hasn’t officially released pricing details for CX Enterprise, although reports suggest it will apply an outcome-based pricing model.

My take

The new AI tools look impressive, but pricing will play a part in their success, as will the readiness of other businesses to set up and take advantage of these kinds of tools.

Disqus Comments Loading...