AI-led production saves in-house and agency teams hundreds of hours, early tests show
Gen AI usage among creative teams is moving from the experimental stage into the everyday. Indie agencies and in-house teams alike are finding major production savings.
IBM’s latest campaign highlights the mistakes AI can make / IBM still from a campaign
Tech company IBM’s latest brand campaign turned the Las Vegas Sphere into an LED fishbowl featuring a school of malformed fish made good by watsonx, the tech company’s enterprise AI product. ‘Trust What You Create’ aims to extol the computer-maker’s artificial intelligence expertise, but also provides an interesting case study of the production efficiencies available to in-house teams using AI.
IBM’s in-house team, along with agency Ogilvy and production partner Mirada, used Adobe Firefly to work up initial versions of each aquatic character and then hired a professional CG artist to create the finished versions. The methodology saved significant time, cutting the production of each character down from 15 working days to two.
The fish are “fantastical creatures that only gen AI could create,” Billy Seabrook, global chief design officer at IBM Consulting tells The Drum, each bearing deviations (such as a body made from mirrored scales) that reference the telltale signs of a gen AI image tool gone wrong. In the face of drift, bias and ‘hallucinations’, Watson X’s rescue reinforces the idea that IBM can steer companies right.
For Seabrook, the time saved during production can be reinvested back into other areas of IBM’s marketing engine. That might mean more time is available to work out the campaign’s strategy, or its media approach.
“If your production timeline goes from six weeks down to two, you can get on to the next campaign and you can drive more personalization,” he says. “You’ve basically given yourself a runway where you can spend more time on the upfront creative concept and not get pinched by your production timeline.”
Firefly is the only third-party AI tool that IBM has been using on commercial work. The workflow used by its in-house team is similar to approaches its consulting arm is recommending clients take, Seabrook says.
Versioning, the practice of making multiple variants of an advertising asset for different audiences and markets, previously took up 80% of the in-house team’s time. “Firefly could basically eliminate that completely,” he says.
“That equates to $100m in savings for IBM. That’s a pretty impressive stat, it’s over the course of three years, but when you start to offload that versioning, localization work, and focus on your hero assets… you’re creating more headroom for creatives who can do more work on campaigns.”
Putting this kind of workflow together for an actual campaign is not necessarily prohibitively expensive, or exclusive to the likes of IBM, one of the original pioneers of AI technology. Kyiv agency Bickerstaff.734’s recent campaign for healthcare brand Esculab involved the use of Stable Diffusion and procedural design tool TouchDesigner to create dozens of design assets, including animations.
The agency’s staff calculated the use of tech saved roughly 80 working days, by measuring the speed of asset production against its previous productivity measures.
Liza Popova, a graphic designer at the agency, told The Drum: “It would have been mega-complicated and time-consuming to draw everything.
“Yes, we had to spend some time getting to grips with TouchDesigner and making it work, making it respond to movements in the video, and ensuring that the text moved along with the lines, but the result is worth it.”
To calculate the time saved, the agency estimated that each design asset for the campaign – of which there were 80 – would typically take a full working day to produce. By contrast, the workflow built around Stable Diffusion cut that time down to 30 minutes. The process of comparing workflows, then, rested upon measurement the agency would have already been carrying out in order to monitor its costs.
Aside from time saved during production, IBM’s in-house team has also been measuring the impact of AI-created assets deployed on real campaigns. Though the work shown on the Sphere was created by a human designer, a shorter campaign produced for a B2B event last year saw Firefly assets used without intervention.
For ‘Let’s Create’, IBM used a single primary design – a question mark – and iterated upon it, using Firefly to create 1,000 variations featuring different typographic textures, copy lines and background imagery for social platform assets.
Like ‘Trust What You Create’, that methodology saved time – taking just three weeks to put together, according to Seabrook – but also improved upon previous paid social performance.
“It was our highest-engaged social media campaign of the year,” he says. “It was very encouraging to see such a good response, and to see the cost savings.”
According to Ari Sheinkin, vice-president of global demand at IBM, the campaign drove 26 times more traffic to IBM.com than previous paid social work.
“Organizations are under incredible pressure to deliver highly personalized experiences across many different channels, and generative AI provides us a path to effectively scale these efforts,” says Sheinkin.
“Adobe Firefly is enabling our marketing and creative teams to collaborate more effectively by streamlining the process of generating high-quality content, which is then used in large-scale personalization campaigns that deliver tailored experiences for customers,” he adds.