The Next Thing – Issue 172

Edit: So when I posted this as a LinkedIn newsletter (it has been on MailChimp for over a decade) I made the comment "I'm not one for walled gardens but I'll give it a shot". It turns out my mistrust of platforms obsessed with optimisation was well founded, as LinkedIn went and spammed every contact I have with a notification or email that this newsletter exists, without notifying me that is was going to do that.

I realise only a small portion of people will actually see this, but I'm sorry for the spam. ??

AI is Ad-tech now ??

Now that we're a solid six months into the latest AI spring, the big digital advertising players are starting to make their moves. In the last couple weeks we've seen announcements from Google, Facebook and Microsoft (currently front-runner for the "Steven Bradbury award for unexpected win in the search wars").

Google's plan focuses heavily on search ads, looking to incorporate generative AI into its Performance Max offering that automates placement, budget and copy. Advertisers will provide various bits of creative, and the AI will remix ads to optimise for performance.

If this sounds similar to Facebook's Advantage+ product, that's because it is (but arguably with far better AI powering it). Facebook exec Andrew Bosworth has also been assuring advertisers that the Meta ad-stack will be bringing all the AI as well, although he wasn't terribly specific about details.

Meanwhile Microsoft is already testing search ads that are integrated into Bing Chat responses. Currently the ads aren't labelled unless the user hovers (not a great start), and Microsoft is floating the idea of image and video ads appearing under chats as well.

If nothing else, it's certainly going to be an interesting couple of years for brands and agencies dealing with this rapid innovation. There's a pretty strong argument that poorly thought out advertising models got the internet into the binfire it's currently in, so it's hard to not be a bit worried about all of the announcements above.

The handing off of creative duties to AI-powered remix bots also highlights a divide that seems to be growing in the broader world of marketing between product companies and brand companies. So much of the ad-tech (now powered by AI) is focused purely on performance, but with zero acknowledgement of long-term brand building. Which is fine if you just want to sell your DTC mattress or juice cleanse. But there's a big question of the effectiveness of any of this when it comes to building long-term brand value. If you haven't read the excellent WARC paper "The best ad for one, is the best ad for all" , I'd highly recommend giving it a look before jumping into the AI-driven creative hype pool.

Not so fast, robot overlords ?

While a multitude of people are trying to work out how to turn AI into advertising dollars, the AI companies are in a high-stakes race to build better and better models. But with all of the attention on these companies, questions and challenges around the data used in training these models are starting to get more pressing.

Universal Music has asked streaming platforms to block any AI companies from scraping both the audio and metadata from their songs. And tech Q&A site Stack Overflow has joined Reddit in charging AI developers to access its data (for those outside the bubble, AI is very very good at writing code, and quite a lot of that capability has come from scraping Stack Overflow questions and answers).

On the media side of things, Washington Post analysed the sources of data used by ChatGPT in building its AI language models. Half of the top ten websites are publishers (NYTimes, LA Times, The Guardian, Forbes and Huffpost), with WaPo itself coming in at number 11. The analysis also showed that quite a lot of data was taken from bottom-of-the-internet sites like 4Chan, and what was once the world's biggest collection of pirate eBooks.

There's two interesting aspects to all this. Current language models managed to get trained on a lot of content that probably should have been charged for. Now the scale of the AI economy is clear, expect anyone making trustworthy content at scale to be charging handsomely for access to their data for training. AI tools are already not super cheap to run - I can easily burn through $10 of API access in a couple minutes building prototypes on OpenAI. So while AI holds a lot of promise for improving productivity, it seems it may come at a cost.

The second aspect that's worth considering for brands is the use of AI generated content in any public-facing way. We're certainly not close to an answer on what copyright looks like in this space, and I'm getting strong flashbacks to the early days of smartphone cameras and social networks, when brands thought it was totally fine to put Flickr photos up on billboards without asking.

Everybody is retail media now ??

After looking on from the sidelines while online and offline retailers built up advertising juggernauts, tech companies are starting to get in on the game. Facebook has been testing a program where advertisers can access retail media networks via its ad platform - offering better targeting for advertisers and increased reach for the retail networks. The pilot has been running in the US with Walgreen's and Dollar General, with results being positive so far.

Meanwhile, Uber has quietly built up a $500M ad business by leveraging first-party data on its customers. If someone jumps in a car to head to a specific store, the Uber app can now show product ads for items at that store. In Uber Eats, advertisers like Pepsi are now showing ads for products after you make an order, with their products able to be picked up from local stores as part of that order.

For all the hand-wringing about the end of cookies and Apple's ATT changes, what we're seeing now is some innovation in the use of data by marketers. Not being able to rely on (usually crappy) third-party data has not caused the predicted ad-pocalypse, it seems to be resulting in smarter use of first-party data, quality partnerships, and most likely a better customer experience.

Not so fast, tech overlords ????

Meta this week underwent its second round of layoffs in six months. Around 4,000 jobs were cut, adding to the 10,000 in November, with another 10,000 reportedly planned as soon as May. The cuts were across a wide range of teams and come amid reports of employee morale being extremely low as Zuckerberg executes his “year of efficiency” strategy.

Meanwhile, over at Google, Samsung is causing a $3B headache by threatening to switch the default search engine on its devices over to Microsoft’s Bing (another point in the Steve Bradbury award tally). While the bulk of Google’s $162B of search revenue comes from paid ads, a large chunk also comes from deals to be the default search on smartphones. With Microsoft aligned to OpenAI, currently a clear leader in text-based AI, Google’s search dominance is under threat.

Only a few years ago, it was hard to imagine a world in which Google and Facebook weren’t the big dogs of the tech sector. But now, a perfect storm of interest rate increases and AI breakthroughs has hit. Burning money is no longer cool, and both companies have made strategic blunders - Google becoming complacent and not innovating in search, while Zuckerberg chased his white whale in the Metaverse.

The TikTok section (it's been a while!) ???

I feel like we're overdue for a TikTok update, despite it sort of being more of the same. Australia's attorney-general has now followed the US, Canada, UK and New Zealand in banning TikTok on government devices . This was "acting on advice from intelligence and security agencies". I've mentioned it before, but banning one app is not doing anything to protect people's data from nefarious state-sponsored snooping. Particularly while our insurers, telcos and banks are doing such a cracking job of publicly sharing customer data.

It's also highly likely that we’ll be lining up to the sequel of this movie, because the most talked about app of the next twelve months will be Lemon8 . Originally launched in Japan in 2020, the social platform has now launched in the US and has quickly become a top downloaded app. Its owner? TikTok's ByteDance.

Advertisers also seem less worried about the concerns of intelligence agencies, with US ad spending on TikTok up 11% in March . Media companies also seem unphased. A recent Fairfax article pointed out that every major news outlet is using TikTok. The only holdout was Sky News, which arguably isn't quite in the demographic sweet spot of TikTok (and arguably is also not quite a news outlet).