AI’s black box problem has clients cautious but curious, agency execs say

AI’s black box problem has clients cautious but curious, agency execs say

On the heels of OpenAI’s ChatGPT emergence, generative artificial intelligence has become the industry’s latest shiny new technology that has captivated advertisers. At least 71% of agencies have taken AI and run with it, leveraging things like chatbots and language modeling for internal processes.

Clients, however, haven’t exactly been clamoring to get their hands on AI solutions, agency executives say. 

“There are very, very few clients out there that are wanting to lead the pack,” said Mark Singer, U.S. chief marketing officer at Deloitte Digital. “They’re all sort of going, ‘This is super interesting. Let me see what it does’.”

AI’s black box problem — in that humans have yet to understand exactly how AI systems compute and make their decisions — may be to blame as advertisers are not yet familiar with the technical nuances that come along with the technological developments.

But it sure is buzzy. AI was the buzzword of this year’s Cannes Lions Festival of Creativity and last year AI-enabled ad spend was said to reach $370 billion, according to Statista data platform. And while many execs agreed it will have a place in the future of advertising, none seemed to agree on what exactly that looks like. The U.S. government only recently announced AI policies this May. 

Artificial intelligence itself isn’t new, dating all the way back to the 1950s. Omnicom Group’s machine learning efforts and business applications go back the last decade. Similarly, Deloitte Digital has made a number of AI-related acquisitions, including Magnetic Media Online’s artificial intelligence platform business in 2018. However, the introduction of ChatGPT last November has launched a generative AI arms race amongst OpenAI’s ChatGPT, Google Bard and others — and piqued advertiser’s interest. 

Interest but trepidation

While agencies are pitching, executives say clients are hesitant to fully commit, interested more in educational information than product implementation, making for a lot of handholding from agencies

“We’re still standing on the shoulders of giants in places,” said Brian Yamada, chief innovation officer at VMLY&R marketing agency. “We still have to be very careful and understand how to best mitigate some of those other risks when we don’t have the luxury of transparency end-to-end.” 

VMLY&R has worked with brands like Wendy’s, Colgate and Starbucks. More than half of its current client roster has expressed an interest in AI, but mostly in the span of education on what exactly AI can do. It’s a similar story at the three other agencies Digiday spoke with for this story, including Ocean Media, Crispin Porter + Bogusky and Deloitte Digital. 

“There’s a lot of interesting potential. There’s a lot of risk. And there’s a lot of understanding,” Singer said. “I don’t believe that, at the end of the day, anybody really understands what to do with it or how it’s going to work.” 

He added that AI-powered tools, like AI image or content generation, has excited consumers, but marketers and advertisers are held to different standards than everyday people, forced to consider regulatory standards, data privacy and content ownership. Finally, AI has yet to prove itself the end all, be all of ad tools with agencies still prioritizing human oversight and intervention.

Thus far, Deloitte Digital teams are testing client-facing technologies like digital twinning, or a process in which a physical, real world activity is replicated digitally, in a virtual environment. For example, if a Deloitte client wanted to know how an ad will perform in front of a certain targeted demographic, depending on location and weather, digital twinning could virtually replicate and hypothesize.

“You’re able to make provocative decisions with a higher level of confidence because you’ve simulated the activity already,” Singer said.

Meanwhile, other agencies are using AI to more quickly process A/B tests, otherwise known as a way to compare two versions of something to determine which performs best, copy editing and visual asset generation, data processing and more.

Per the executives, costs associated with testing these tools are in paying for AI services, like Bard or ChatGPT. (No executives offered further details around AI ad spend.)

Education is key

Transparency around what data is being used to train machine learning and AI tools, what AI-powered tools are being used and how the agency is using them has become crucial in helping clients understand what AI can do, said agency executives. 

In addition to human oversight, agency executives say they’re protective of what data goes into these AI-powered tools to ensure understanding of the source material the machines are learning from. 

At Ocean Media, the analytics team doesn’t store any personally identifiable information to train AI models, according to Annmarie Turpin, chief technology officer for Ocean Media’s analytics team. VMLY&R uses a process called mapping, walking clients through the process, from what data points are being entered into the machine to what software programs are interacting with it. 

While clients are curious, they’re not as apprehensive about AI as they were years ago, when AI “started as this catchall for things that were cool,” said Turpin.

For clients who are willing to leverage AI tools, she suspects they may be more comfortable given technological advancements or are simply more interested in the results than the complexity of its processes. 

“I feel like I have more of a license to work in black boxes today than I did five to 10 years ago,” Turpin said. “It can do a lot of things in a little bit of time.”

https://digiday.com/?p=513835

Read More

Leave a Reply

Your email address will not be published. Required fields are marked *