Gen AI for the Media Industry: Two Years On

I was looking at a report I wrote in 2023 on how media and publishing companies should respond to the emergence of Web 3 and Gen AI (Link), and much to my surprise, it is still very much current and applicable for the foreseeable future.

Given how much Gen AI has evolved since the launch of ChatGPT in November 2022, industries, especially well-established ones like media, saw no significant impact from the hottest technology on earth (literary). There have been no disruptive transformations. The regulatory challenges, customer attitudes towards Gen AI, the nascent technology, and high global capital costs may have all contributed to the slow adoption. Gen AI is, after all, only moving out of the Trough of Disillusionment, according to the Gartner Hyper Cycle. It fits any product development or evolution curves, for that matter. However, re-reading my report reaffirmed my view: the main challenge is the non-existing strategic alignment with the emerging tech landscape, slow response to the fast-shifting societal behaviours, and a lack of innovation mindset and drive to harness the opportunities.

Change is hard, full stop. Businesses are naturally cyclic. According to the London Business School (2018), the average lifespan of a company is only 17 years. In the digital world, it is even shorter, as we know that innovation and diversification are the only ways to avoid the death of a frog in a slow-heating pot. Large companies, with their constant operational KPIs and P&L pressure, often have little attention left for innovation. Leaders frequently say the company must embrace innovation in order to thrive but do not always act on their promises. Waiting for others to work out the path ahead is a valid strategy, but it has an opportunity cost. Unfortunately, the entire media industry constantly plays the follower role. Let the Googles, TikToks, and OpenAIs fight; we'll go with whichever platform the audience congregates on and get a scrap for our content. Let's hope the government does not favour big tech and lets AI companies use our content for free. Oh wait, they already did.

What is happening out there?

Sentential statements, such as 'AI Will Replace 95% of Creative Marketing Work' (Sorry, Sam), have long lost their click-bait lure. Many are still trying to figure out that one single killer use case that will upend the world - like what PCs did to office workers or Web 2.0 to e-commerce. So far, no one has cracked it, yet. Instead, we saw a build-up of knowledge, real-world experiences, and intuition about where and how to adopt and adapt and to see the immediate value. Tools and frameworks, both technical and regulatory, are slowly emerging. For example, many have gone agentic in the past year and the passing of the EU AI Act. Visionaries quietly built new products, such as Perplexity AI, that grew into formidable players rapidly.

Influential players in the industry, such as The New York Post and NewsCorp, first sued AI companies for copyright infringement and then quickly made deals with the accused to allow their content to be used for training LLMs. Smaller players followed suit, and many partnered with big tech to serve news stories via AI products and platforms developed by the tech giants. It seemed a win-win, and everyone was happy.

Two things are going on here. First, it is common for established industries to defer disruptions and potential dangers through lobbying and regulatory levers. This is precisely what they did. Second, this is not a surprising power shift. AI companies are barely building platforms at this stage, are not profitable, and have not found a way to encroach on other industries' gold mines. ChatGPT is not a killer product that is capable of deferring new entries. Google, Anthropic, Mistral, and DeepSeek have proven that if you put enough money in, you can build an LLM that is as good as OpenAI GTP, with far lower costs in some instances. The barrier to entry is low. ChatGPT subscription will not make enough money to sustain OpenAI - the $200 monthly ChatGPT Pro subscription indicates how much they will have to charge to make it a viable business model. And, with the release of Claude.ai, Mistral Le Chat and many more, the price war has only just started. Price wars are a losing game for the entire industry, and everyone loses.

So, AI companies have built the platforms and enticed enough attention; what next? Monetisation, of course. But how - if chatbots are not the answer? A new trend has emerged, which should not surprise anyone. Have you seen the ads in Google AI Overview and ChatGPT? Yes, that is right - digital advertising always comes to the rescue - for anyone with access to a large audience. Is this any different from how Facebook, Instagram, or TikTok started? Of course not. Until AGI becomes a reality, this might be the only viable business model. AI companies are racing to AGI. Capability is one thing, but hallucinations and errors are another. "Better", say from 90% to 96%, is not good enough. Many use cases require 100% accuracy, but would Gen AI ever get there? The road is unclear.

Within established industries, companies have tried to optimise or automate workflows with LLMs, find ways to make their products AI-enabled, or use Gen AI as an RP opportunity to enhance their brand images. Is this the natural approach to embracing innovation?

Let's step back and look at the outputs: regular chat boxes, search-the-news-archive with RAG, upload documents and ask questions, transcribe recordings and meetings, and so on and on. Let's scrutinise the efforts from a value-creation point of view to see if there is anything that individual firms can use to enhance their competitive advantages. We find nothing. There is no differentiation, no enhancement for existing advantages or real integrations. The blanket 'efficiency gain' is often useless for real KPIs when not aligned to firms' strategies.

It felt that everything was still in the tech feasibility phase; worse, everyone was reinventing the wheel. This may seem logical for adopting new technology, but what about buy-vs-build? Do you know that Google Agentspance and ChatGPT Enterprise can do them all? If you want a secure, more flexible solution, LibreChat can get you up and running in hours. What is missing here is the strategic alignment and integration of Gen AI with firms' proprietary data and existing business models. AI tools are ubiquitous, and you are building one, too. Are you in the AI tool business now? If not, stop trying to tick that 'We are AI now' box by building yet another undifferentiated RAG solution. Upskilling workforce aside, being able to upload private slide decks and spreadsheets is not a valid justification for building your own ChatGPT clone.

Finally, let's pick three headlines from last week. First, newspaper circulation numbers are low in the US, with The Wall Street Journal leading the chart at around 460k. Second, magazines on Apple News are paid peanuts for their content. But it's okay. A little extra is better than nothing. The trouble is that they don't know who bought the content; they do not know their potential customers because Apple owns that relationship. Third, GB News has made a smaller loss than last year, and they have more online audience than TV viewers! The last one, Reach (the media company) had an excellent year, but the market cap is tiny in proportion to its revenue. It is all because they are mainly in print, not online.

Disclaimer

The linked report does not contain any company's private information. All data was obtained from the Internet. The imaginary Company X may resemble any big UK publisher, such as Reach, dmg::media, News UK, etc. It is solely based on readily available information on the Internet and the authors' knowledge and experience in the media industry.

References: