The impact of AI on the media and advertising industries

Talking Heads with Matthijs Leendertse

Daniel Morris: Hello, and welcome to the BNP Paribas Asset Management Talking Heads podcast. Every week, Talking Heads will bring you in-depth insights and analysis on the topics that really matter to investors. In this episode, we’ll be discussing the impact of AI (artificial intelligence) on the media and advertising industries. I’m Daniel Morris, Chief Market Strategist, and I’m joined today by Matthijs Leendertse, Senior Lecturer for Media Economics at Erasmus University, Rotterdam. Welcome, Matthijs, and thanks for joining me.

Matthijs Leendertse:  Thanks for having me, Daniel.

DM: When we think about the impact of AI, the first things that come to our mind are the impact on employment. Are we all going to lose our jobs and be replaced by AI agents? Interestingly, we seem to have seen the biggest impact so far on the technology industry – big tech companies laying off workers to invest more and more in AI – arguably not so much on other industries, and probably more as individuals or as employees.

We see how we’re using AI, but perhaps don’t perceive the impact it is having across different industries, in particular, advertising and media. Perhaps you can give us some insights on that. Let’s start off with advertising. Historically, creative agencies have captured their highest margins on high-volume execution, scaling assets across channels, rather than strategy itself.  As generative AI automates that production layer, how do agency revenue models need to evolve, and what does this mean for the traditional billable hour?

ML:  When most people think of advertising, they tend to think of the Mad Men era, where you had larger-than-life creatives who created iconic slogans and had big strategic ideas. But if you look at the actual operational engine of traditional agencies, the bread and butter was always the heavy lifting of execution – generating dozens of campaign variations, localising assets and managing high volume junior labour hours.

But what we are seeing now is what you might call a ‘madman in distress’ era, because generative AI hits that execution layer head-on; edtech and generative tools are so accessible that clients themselves are increasingly using AI to handle execution, preliminary ideation, and asset variation in-house. Tasks that used to take an agency weeks of production time can now be done almost instantly on the client side, and that inherently puts massive pressure on traditional billable hour models.

However, it doesn’t mean agencies will disappear. Successful agencies are finding new life as what you could call strategic guides – helping clients navigate and pivot into the AI era, and to adapt on the technology side. You see a lot of agencies, particularly the bigger ones, are aggressively building and buying AI capabilities to guide their clients through this transition.

Take Publicis Group, for instance, the big advertising powerhouse from France. They are acquiring AI content intelligence platforms like Edge AI to give brands predictive, real-time analytics on what content actually works. So instead of just selling hours to execute assets, forward-looking agencies are developing enterprise AI engines and advisory tools to help clients orchestrate their own marketing tech stacks.

And technology is not the only thing; it’s only half of the equation. There’s also a critical human component that software alone cannot replicate, because when algorithms can generate endless creative variations, the real bottleneck actually becomes human judgement, curation, and particularly brand consistency: you can personalise campaigns all you want with AI, but at the same time, a strong brand often relies fundamentally on its shared cultural meaning – the collective story that everyone recognises and makes the brand interesting for consumers.

So, I believe agencies that successfully integrate this deep technological AI capability with human judgments and brand stewardship will be the ones that thrive in this new world.

DM: Let’s move on to media. If we start with digital media and video, I believe we’re seeing two major shifts happening at once: A complete transformation in how content is produced, and a move towards real-time algorithmic delivery. From a sector perspective, how is AI reshaping the economics of video production, and what are the broader retention and monetization implications of these hyper-personalised content loops?

ML:  We are seeing disruption hit both the supply and the demand side of media. If you look at the production side, what we see is that AI is drastically lowering the cost and timeline, especially of video creation in studios and newsrooms. We are seeing what you could call a centaur model emerge, where half man, half machine work on research rendering and automated editing, and it frees creators to focus on the really high-level storytelling.

But it also lowers the barrier to entry so drastically that the sheer volume of video output and other types of output is exploding. To give you a sense of the scale of this, this week alone, Spotify removed 75 million AI-generated songs from its platform. What’s remarkable in media production is that because of AI, the friction between an initial idea and its final realisation is more or less disappearing, and it’s very similar to the concept of vibe coding that we’ve seen in software development, where you actually don’t need to write code, you just express in your own words [your] intent or vibe, and the AI handles the execution.

In media, you no longer always need complex production gear, camera crews, or advanced rendering software to turn a high-level creative vision into a broadcast-quality visual asset, and this collapse in production friction leads straight into the distribution side, where content and distribution effectively merge.

We are moving from an era of the ‘infinite scroll’ – where algorithms select these pre-existing clips for you, like on TikTok and Instagram – to ‘infinite creation’, where the distribution engine itself generates or adapts media in real time, tailor-made just for you.

We already see that viewers are hooked on TikTok curating flat videos for them. But imagine when the content becomes hyper-personalised, where video streams, 3D environments, or narratives dynamically mutate based on real-time micro gesture-tracking and other forms of data.

Behavioural scientists that study games, etc. are already using the term ‘digital heroin’ for personalised content streams because they are so addictive. But can you imagine how addictive media becomes when it isn’t just curated for you but generated in real time, tailor-made just for you?

And this raises profound regulatory questions about public health and well-being, as we are already witnessing the damaging mental health effects of social media and gaming addiction. This could take that to a new level.

But to go back to the economic story, the broader economic paradox strikes at the core funding model of the entire creative and information ecosystem. Primary creators face a direct displacement effect – where cheap automated synthetic output begins to replace the work human creators make – although its success depends on user acceptance, something I’m now researching with my Master students.

But the structural thread goes deeper. If we replace and starve human creators before even understanding the long-term implications, we risk undermining not only the cultural ecosystem that feeds these AI systems, but also the essential democratic function of the media, which depends on a financially viable, independent human press to inform the public and move power to account.

DM:  Well, I have to admit, some of that sounds quite scary, but perhaps I’m simply too old to appreciate the benefits of all of this! Let’s end on regulation, Matthijs. There, the European Union recently enacted the EU AI Act to address some of these systemic risks you’ve mentioned – deepfakes and algorithmic governance. From an institutional perspective, does this framework adequately address the structural risks facing democratic information ecosystems, or are there still underlying economic blind spots?

ML:  I believe the EU AI Act is a groundbreaking piece of regulation, but we also need to be clear what it actually does and where it falls short.

At its core, a risk-based framework for AI is absolutely necessary. It gives us vital tools to combat large-scale disinformation campaigns, which we have seen all over the Western world; to mandate transparency for synthetic media like deepfakes, so everything is labelled and people know it has not been created by humans; and to ban manipulative algorithms that threaten the very integrity of democratic deliberation.

With this EU AI Act, we have to classify systems – for instance, those used to influence elections or alter voter behaviour – as high risks, so that establishes essential safeguards for our democracy.

And big tech, of course, has aggressively lobbied against this, precisely because it shifts liability onto model developers. It imposes heavy compliance burdens on foundational models, and it limits their ability to deploy un-curated algorithms in the European market. From a safety and democratic defence perspective, it sets a global gold standard. However, from a macro and a media economics view, the act has two major omissions, in my view.

First, it regulates safety, but it ignores value distribution. While there is separate EU digital legislation, the AI Act doesn’t solve the underlying economic extraction, where platforms scrape original journalism and content without compensating the creators. Starving the financial engine of independent media remains an unaddressed risk to long-term democratic accountability.

Second, if Europe wants sovereign AI that reflects our public values and democratic standards, regulation alone is not a strategy. You cannot regulate your way to competitiveness and digital sovereignty without economic investment, massive capital deployment and infrastructure incentives. Europe risks total dependency on foreign – particularly US and Chinese – AI stacks.

We do see promising European champions emerge – Mistral in France with its foundation models, Lovable in Sweden pioneering five coding, or German company, Helsing, in AI defence technology. These companies prove that European innovation can compete at a world-class level.

But, to maintain a sovereign democratic digital ecosystem, policymakers must [bring to] bear regulatory guardrails against disinformation with aggressive economic investment, ensuring European AI is built on European values rather than purely imported.

DM:  Thank you, Matthijs. Some of the key points that you mentioned: first, the significant impact that AI has had on the media and advertising industries, causing companies, agencies, to pivot away from just generating billable hours to higher value-added content and the need for human intervention and judgement. You see key developments in the hyper-personalisation of content and mention some of the risks that that may entail.

In response to that, we do have the new EU AI Act, which you noted seems to be a good first attempt to manage some of these risks, but you pointed out it is not sufficient to enable Europe to manage or control the development of AI and its impact on society in the future. Matthijs, thank you very much for joining me.

ML: Thank you very much for having me.

DM:  That’s it for this week’s episode of Talking Heads. If you would like more information about our capabilities in investing in the AI theme, please reach out to your asset management contact or check out Viewpoint, our website for investment insights at Viewpoint.bnpparibas-am.com.

The Talking Heads podcast is available on Spotify and on YouTube, and this is real human-generated content – it is not AI! For YouTube, visit youtube.com/bnppam/playlist and tap or click on Talking Heads.

You’ve been listening to the BNP Paribas Asset Management Talking Heads podcast with me, Daniel Morris, and Matthijs Leendertse, Senior Lecturer for Media Economics at Erasmus University in Rotterdam.

Please do join me next week. Until then, take care.