Adobe Sneaks: AI for Generating Headlines and Teaser Images

Hiker looks at cliffs by the water.

Once a year, a call goes out to thousands of employees in Adobe offices around the world. Anyone in the company, from engineers and data scientists, to UX designers and product managers, have a chance to put forth new and innovative ideas that evolve the way brands engage consumers digitally.

Over time, Adobe Sneaks have become an important innovation engine for Adobe, working its way into our apps and delivering popular features such as an AI assistant in Adobe Analytics and AI-driven asset tagging in Adobe Experience Manager. The submissions this year, focused on AI and mixed reality technologies, were whittled down to a final set of 7 projects that we share publicly. And while these projects are not yet available, around 60 percent of Sneaks eventually make it into an Adobe product.

As part of our ongoing series, we are sharing details here on a new project that taps AI for content creators.

“Project Snippets”

As more people stay home, more eyeballs are going online. Content creators are seeing record spikes in traffic, with people consuming more news and discovering stories that companies tell on their owned channels. The Adobe blog for instance, saw a 30% increase in April traffic (over March), with content focused on helping our communities stay productive and powering digital business.

As the “new normal” begins to take shape, the move online is expected to persist; Consumers are getting more accustomed to opening up a browser or app for all of life’s needs. And as a result, brands that were already under pressure to deliver more content are having to do even more. We are showcasing “Project Snippets” in Adobe Experience Manager to show how AI can assist content creators. It is meant to accomplish two seemingly opposing tasks: do a bit less work, but drive more engagement.

To best illustrate this, imagine a tourism company, which we will refer to as ‘Wknd’. They are gearing up for the moment when travel begins to pick back up. Their content team has produced a lengthy piece, around all the things that people can do in Australia. Knowing that their readers tend to choose stories by scanning headlines and images, they start to work on a teaser.

Project Snippets uses Adobe Sensei, Adobe’s AI and machine learning technology, to instantly generate the assets via natural language processing and computer vision. It suggests a relevant header image, headline copy and preview blurb (saving the team a bit of grunt work). And that is just the beginning; Wknd also knows that people have different backgrounds and interests. It wants to personalize the teasers, for a set of audiences that include thrillseekers, frugal vacationers, upscale luxury travelers, wildlife enthusiasts and nightlife party goers. Adobe Sensei will help deliver variations based on copy and image qualities that perform best with each of these segments.

Rock climber image on a story about Australian adrenaline junkies.

The image itself can be pulled from the brand’s own repository, or through services like Adobe Stock. Above and below, you can see what AI-derived images have been selected for the thrillseeker audience and upscale luxury audience, respectively.

Vineyards image on a story about Australian wine.

The suggestions can be edited by the user as well. If the team does not like the initial recommendation, Project Snippets allows them to click through a set of alternate image and copy options provided by AI. The “Snippet Quality Score” (93% as shown above) is provided as a click-down for the user, with context on how Adobe Sensei arrived at its decision. From a personalization standpoint, it can show which high-performing image and copy qualities were used as a benchmark.

The Wknd team now has 5 unique AI-generated teasers to support the article on things to do in Australia. They can set up the experience where the teaser changes depending on who is visiting the site. For some brands, this can be implemented based on subscribers who have set up a user profile. In other cases, teams are able to capture signals from previous site interactions, to make a determination on what would be most engaging for the visitor.

With Project Snippets, the goal is to show how AI can be a helping hand. The final decision is ultimately up to a human, but the machine interaction could be a symbiotic one. In this case, AI is taking on the more laborious work of having to find an image and write copy, but also helping marketing teams boost engagement with better personalization. It can be a welcomed assistant for many teams that are now needing to do more, with the same or less resources.

At Adobe, we believe that everyone deserves respect and equal treatment, and we also stand with the Black community against hate, intolerance and racism. We will continue to support, elevate, and amplify diverse voices through our community of employees, creatives, customers and partners. We believe Adobe has a responsibility to drive change and ensure that every individual feels a sense of belonging and inclusion. We must stand up and speak out against racial inequality and injustice. Read more about the actions we’re taking to make lasting change inside and outside of our company.

We also know many people are still impacted by the current COVID-19 crisis and our thoughts are with you. The entire Adobe team wants to thank you, our customers, and all creators around the world for the work you do to keep us inspired during this difficult time.