Mark Zuckerberg announced this week that Meta now has a team dedicated to creating artificial intelligence-based tools. It plans to bring text, image and video generation technologies to WhatsApp, Messenger and Instagram. It could be talking about chatbots, creative filters and unusual advertising formats. Almost simultaneously, Insider Intelligence examined Meta's existing experience of incorporating AI into its advertising technology and found that there are those among advertisers who are not happy with such tools.
Insider Intelligence's research sums up the first results of Advantage+. This AI tool from Meta can automatically generate up to 150 combinations of different elements in adverts, depending on the specific objectives of the marketer. Algorithms then check all options and select the most effective, with the ability to automatically change text and images.
An October report on Advantage+ seemed to show cost savings and efficiency: the cost per thousand ad impressions (CPM) was reduced by at least 20% and the cost per click by 15% (Tinuiti data). Clients of marketing agency iProspect, for example, earned $7 for every dollar spent.
But in the face of these encouraging statistics, some agencies have begun to abandon Advantage+. Several things have put clients on guard. First, marketers are frightened by the loss of control over advertising campaigns. Apparently, people aren't yet ready to delegate much of their work to artificial intelligence. Secondly, as one of the advertisers who left Advantage+ noted, "the tool lacks human emotion and common sense": it sometimes generates ads that provoke negative comments, hate speech and calls for violence. Thirdly, advertising and marketing agencies may feel threatened by Advantage+: with such parsimony and efficiency of AI tools, brands may start to wonder whether they even need agencies as an expensive layer between them and Advantage+.
Insider Intelligence reminds us that Meta lost about $10bn in revenue in nine months after Apple unveiled its new privacy policy in April 2021. Meta must now rely more on first-hand data, such as what users like and comment on, whether they mention certain brands. Against this background, analysts remind us that Meta has a history of inflating metrics to its advantage. In addition, the corporation hopes to use the agencies to convince brands that the social giant is a reasonable place to invest in advertising, so it should not complicate its relationship with them by improper Advantage+ work.
At the same time, Meta is working on its own large language model, LLaMA, which is now geared more towards researchers than the general public.
