Hello everyone from London, England, as usual. 

This latest newsletter is, for want of a better description, an “AI Special.” We have suddenly all become conscious of AI in our midst and what it can or can’t do for our news organisations. But there has not, as far as I am aware, been a decent conversation around what its role in media advertising is. 

Is it a good or bad thing? Does it assist or hinder us? How can we use it as a tool for advertisers? And what are they saying about it? All this and more to follow. 

I had a recent meeting of my INMA Advertising Initiative committee where we dedicated much of our time together purely to talk AI and all the things mentioned above. It was an interesting debate, centred around what is happening in the German market, where it seems they are using AI more than most. There are two members of my committee from German media houses who led the debate and whose comments and experiences I am sure you will find intriguing.

Along with the support thoughts from other committee members, from places such as Canada, USA, Brazil, and Ireland, there is what I believe to be a “state of the nation” summary of where we all are today as a media industry. 

A futurist’s take

First, let me set out some of my overall AI observations, particularly from Gerd Leonhard, a German futurist who has been both a keynote speaker at a previous INMA World Congress and a special guest on my UK radio show (The UK Brand Show). 

Gerd addressed the recent Indian media festival in Delhi where he talked AI. AI can and will certainly be transformational, he said, but how it shapes all our futures will be largely determined by what we (as a human race) want from it. He paints picture of heaven or hell.

AI heaven is where it provides efficiency and productivity gains by offloading “commodity work”/shifting routine work to machines and accelerating human knowledge via research and discovery. 

And you can immediately see lots of benefits in advertising for getting rid of commodity and routine work here, be it helping write better pitches, automating sales lead outreach, scheduling of meetings, offering chatbot services for advertisers, tailoring product ads, et al. 

However, Gerd describes AI hell where it brings us much unpredictable bias and error (can it be trusted?), alternative realities based on illusions and simulations, and even de-humanising us all as a race. 

De-humanising? Gerd says he wants his computer to do a competent job of what he tells it to do ... not to have its own sense of what is right or wrong and determining outcomes for us. That could be disastrous in some areas.

He used a chilling example of asking the Indian audience what AI would say if we asked it, “How can we solve the problem of climate change?” Its answer, he says, would be, “Kill all humans,” as “algorithms know the logic of everything but have the feeling of nothing.”

So, maybe there is a cautious approach we need to take when bringing AI into advertising, too? 

The question, I guess is what do we want from AI? I will return to Gerd at the end of this newsletter, but what does the media industry think? I asked my two German committee members to lead the debate.

Advertising Initiative committee members weigh in

First we heard from Thomas Schultz-Homberg, CEO at KStA Media in Cologne, Germany: 

“I see AI as neither heaven nor hell. We are conducting lots of experiments with generative AI for producing content and for rewriting existing content as new stuff for search engines. Lots of work in curating news. The recommendation boxes where we have articles on are not being curated by humans anymore but by a machine learning/AI system. This personalises the recommendation based on the user profile and chooses the main article the person reads. 

“The CTR has increased by 70%. Very encouraging. We did expect an increase but not as high. We do this experiment for a number of reasons: 

• “To do the ‘stupid’ commoditised jobs previously done by humans. We don’t want to replace humans at all, but AI can do some of the basic jobs around curating, publishing … and in advertising. 
• “The power of AI fosters speed in a lot of business modelling and conversion. Specifically in the programmatic area, in some fields, we are experimenting with driving dynamic pricing models. We no longer specify floor pricing. AI takes over the price control. That means we look at the current demands and dynamically adjust the floor price based accordingly in real time. So, if we see a high demand we can raise the floor price, and if vice versa, we can address any advertisers not willing to pay a higher price. Better to get some money than none. 
• “We look at the behaviour of readers as well. The AI doesn’t just look at demand but also reader profile to see if there are people who might be interested in say, a special campaign, and in real time try to connect the right reader profile to the best case, high demand, high priced campaign.”

Sound like voodoo!? 

Thomas continued: 

“Several AI machines combine to make a decision, on price, campaign, to which reader, where it is displayed (in what section, etc). It might sound like voodoo to some, but it works. We see an increase in the number of published campaigns, in the average price we get (this is rising, too), and we are more ‘sold out’ than when we have a static floor pricing.

“The interaction with readers is much higher when we had static personas in the past. We are now looking to increase these experiments. We have always seen increases so far. Some higher than others but always an upward trend.”

How many people are working on this, I asked? “It is all worked on with a mix of sales team and data team, currently about five to eight people are working on the experiments.”

Is it an in-house proprietary, self-build? “We use several tools which we combine together. Some from Google, some from provider, Pubmatic.”

Then Stephan Thurm, chief digital officer at Funke Media, added his thoughts:

“We share the same path as Thomas. Main resources go on publishing supported by editorial. There are reservations about AI, but I am convinced here are more opportunities than threats but that changes depending on which part of the company you look at. 

“Our strategy it to look at commercially available tools and combine them. We have a strong data team who are able to do the job of combining. 

“We did develop two things ourselves: 

• “Build a system called Noah. It’s for emergency issue production. They idea is how do we produce a newspaper if everything is ‘down,’ maybe due to a hacker attack? We felt we needed a system to take on content, however produced, and turn it into an emergency newspaper for the next day. We trained the system with hundreds of thousands of articles. It’s not perfect but it’s pretty good.
• “Then we plugged in different sources, into a huge content hub. The machine can decide which articles belongs in which publication we have. It can also re-write or short stories, ‘for example, it can suggest headlines for the editors.’

“From the above we produce new products in various verticals … which makes it all very interesting.”

But this of course is all about publishing. So, where does advertising fit in, I asked?

“We have developed a system that deals with contextual ads. With skim links, we can put advertisers behind them. We developed this ourselves and put machine learning/AI in the process to deliver better results for advertisers, so the contextual power is better. Various commercially available tools help support this engine. We made it easy by integrating it into Google Ad Manager. So, we don’t have a second system for this, it can all be done in Google Ad Manager.”

Contextual advertising opportunities

“The next goal is to use it for SEO and the internal linking of our products. We have about 120 ad products, and the machine can undertake contextual linking with those products. It’s an important use case. It’s all pretty interesting in relation to contextual advertising. 

“Whilst we built our engine ourselves, it’s important to point out there are many commercial tools as mentioned upfront that can help us in our quest. There are many services you can buy to supplement your own efforts. For example, DMPs, ad servers (we have ad servers that are AI backed for video). So, we have a lot of commercial tools, which, sometimes we combine, sometimes not.

“We also do dynamic floor pricing through a vendor for programmatic advertising, and we do content production for advertisers with all kinds of AI tools for performance stories (ie, those stories written by commercial editors who write about subjects in a way we can commercialise them, in future, maybe via e-commerce, etc.) These stories are paid by performance. AI has helped us increase the output here. Sometimes the AI-backed stories perform better than the handwritten ones (written by our editorial journalists).” 

What are the downsides of AI?

Thomas: “It’s not is easy to control how AI decides [something]. So we are not making a clear path as to how decisions are made. The more complex the AI system is, the less you can detect why it did this or that. There is a loss of control, especially when you use that in real time tools, like advertising markets. This might be a downside as we are not used to this feeling of loss of control.

“But, all we can say is so far, is that everything is showing better figures than before. It’s true that total control is not possible. We don’t know what the outcome would have been if a different path was chosen still using the tools. What would have been the outcome if we could have influenced it? Would it have been better still? We just don’t know.”

Other members of the committee were invited to add any comments they had. Tracy Day, managing director of ad products and innovation at The Globe and Mail in Canada, said: 

“We are starting to use AI for recommendation engines and custom content — not to take over but for directional purposes. It’s really good with its quick speed for things like social posts. Yes, also we identify unsatisfied customers. Our engine shows us when customers quit. If that happens, we get alerts to contact the customer etc. Ultimately it’s about insights, better UX, and more revenue.”

Latha Rao-Cheney, senior VP/MediaNews Group, Detroit, United States, added: “For us, we have been dabbling with decision tress, like Tracy mentioned, with recommendation engines. Simply put, it’s about campaign management for low-tier campaigns where there is minimal risk involved and where it’s far more transactional or writing ad copy.”

Marcelo Benez, chief commercial director at Folha de. S.Paulo in Brazil, said: “We are still defining what’s the best use of AI in different areas. We are currently using AI for briefing and generating new ideas. We see that AI doesn’t replace the human but can work well for certain tasks and activities. The need for human supervision is still a reality.”

Are we ahead of our advertisers when it comes to AI — even if we are at base one currently? 

Marcelo: “We are ahead of our agencies and advertisers. They don’t ask about nor show AI to be an important reality in our market.” 

Elmer Moran, media solutions director at The Irish Times, Dublin, Ireland: “If advertisers are showing promise in this area, we usually hear about it from some of the big advertisers or brands, but we are hearing nothing currently. Not even from ad agencies. Some talk maybe but no one is saying, ‘This is what we are actually doing.’ I feel that publishers are ahead.” 

Tracy: “I agree, but I think the fear from agencies is that clients will take content into their own hands. That is not necessarily the case though as clients wouldn’t necessarily create content that would be right for the audience. For example, would it be edited properly? And they don’t necessarily know if it’s ethical or not, etc. There are a lot of (ill-conceived) issues around that, but I think the biggest fear is that agencies feel that AI can help do it for clients by themselves.”

Stephan: “We talked to the big media agencies about AI and their plans. They say they are not so optimistic. They use it for media planning support, but they still say that’s a ‘human thing.’ They say they have tried to optimise programmatic on their (buyer) side, but I am not hearing they using it for creative.”

Summary 

So, there we have it. I think my committee demonstrates where the global market is at. Some publishers are somewhat “down the road” and some are at base one. Advertisers aren’t really at the races yet, but it seems they would appreciate the help and insights if we can provide them.

Maybe if you’re not yet experimenting with AI, you could try some basic steps first. Maybe get it to draft some consumer surveys and analyse the results? Use it to create ad copy variations or media strategies and see what it comes up with? Could it generate video from text for you? Could it create some audience segmentation for targeting? Or even create some social media posts? 

Try it. Experiment with it. And good luck. You might find that it won’t be perfect, but the question is, will it be “good enough?” The answer to that might scare you. 

I leave you with my friend Gerd Leonhard, the futurist: 

“Technology has no ethics. Embrace it but don’t become it. Don’t be a commodity. Commodity jobs will be robotised if not already, but robots don’t have feelings, emotions, intuition, etc., that humans have. The person with AI will win in the future, but AI will not beat the person. We need the tools but we need to control them.

“The benefits of AI as far as I see it that it must benefit four Ps: people, planet, purpose, prosperity (economy). Take control of your future.”

Food for thought? I think so. 

About this newsletter 

Today’s newsletter is written by Mark Challinor, based in London and lead for the INMA Advertising Initiative. Mark will share research, case studies, and thought leadership on the topic of global news media advertising. Sign up for the newsletter here.

This newsletter is a public face of the Advertising Initiative by INMA, outlined here.

E-mail Mark at Inma.mark@gmail.com with thoughts, suggestions, and questions or follow him on Twitter (@challinor).

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .