When AI Can Make Anything, What Becomes Valuable?

Artificial intelligence inevitably featured heavily at Social Media Conference Cymru, but the most interesting discussion was not about which tool could generate the best image or video. The bigger question was what happens to creative value when the cost of producing content falls dramatically.

That is where Dr Alex Connock’s session became particularly interesting. His argument was that generative AI is making competent content incredibly cheap, fast and scalable. Text, imagery, sound, video and even software can now be produced in volumes that would have been unimaginable only a few years ago.

Humans cannot compete with machines on volume, and perhaps they should not try to.

Connock framed human creative work almost as a luxury product. Human-made work is comparatively slow, costly and difficult to scale, but it also has the potential to contain first-hand experience, judgement, originality and creative leaps that are not simply extrapolated from existing training data.

That is a much more useful way to think about AI than the endless debate about whether it will replace creative people. If average content becomes abundant, the value of average content naturally falls. The scarce things may become originality, access, lived experience, genuine expertise and taste.

In other words, AI does not necessarily make human creativity less valuable. It may make the best human creativity more valuable because it becomes easier to distinguish it from the enormous volume of competent but derivative material around it.

Connock’s strongest idea came when he talked about searching for what historical data cannot yet predict.

“Look for the thing the algorithm doesn’t know.”

His Star Wars example made the point neatly. Before Star Wars existed, there was no historical dataset proving that audiences were waiting for it. Once it succeeded, the industry had evidence that space opera could work and could then produce endless variations on the theme. But the genuinely new idea arrived before the pattern was visible in the data.

That is one of the limitations of optimisation culture more broadly. Algorithms are exceptionally good at telling us what resembles something that has already worked. They are much less useful when the opportunity is something audiences have never encountered before.

The implication is not that marketers should ignore data. It is that data and AI should inform judgement rather than replace it.

Connock was certainly not anti-AI. His tool list covered everything from agentic systems and coding to images, sound and video, and he spoke particularly enthusiastically about Grok Bots during the Q&A. His point was much closer to using the technology aggressively for the work it can do well while being deliberate about where humans still create disproportionate value.

That distinction was echoed by Mared Parry from Betches UK. She spoke about using AI to support areas such as scripting and workflow without asking it to become the personality of the brand. Betches relies heavily on real people, recognisable faces and a tone of voice that feels genuinely human.

That creates a useful model for many brands. AI can sit behind the scenes helping with research, organisation, transcription, ideation, formatting, production and repurposing. The audience-facing value can still come from people, expertise, humour, opinions and real experiences.

Liberty Marketing approached the same shift from the perspective of search and discovery. Their presentation suggested that AI is increasingly playing a role earlier in the decision process, with people using tools such as ChatGPT to explore problems, compare options and discover brands. Traditional search remains highly important, particularly as people move closer to action.

They also discussed traffic coming from ChatGPT converting strongly in some of their data. One possible reason is that the user may already have completed a substantial amount of research before clicking through to the brand’s website.

The most interesting implication is that visibility in AI environments is not simply a question of creating more pages for your own website. Liberty highlighted the wider digital ecosystem around a brand: journalists, backlinks, third-party websites, social conversations, reviews, creators and user-generated content can all contribute to how credible and visible a brand appears online.

This makes the boundaries between PR, search, social, content and brand reputation increasingly difficult to separate.

It also strengthens the case for producing genuinely original material. Liberty discussed the concept of information gain, essentially the value of contributing something new rather than publishing another version of what already exists.

That could become one of the most important implications of generative AI for content marketing. If a model can already create a perfectly adequate introduction to almost any well-documented subject, there is little strategic value in publishing thousands of words that merely repeat the existing consensus.

Original research becomes more valuable. Proprietary data becomes more valuable. First-hand experience becomes more valuable. Good interviews become more valuable. Strong opinions backed by expertise become more valuable.

The irony is that AI may push content marketing back towards the things that made good content valuable in the first place.

Phill Agnew from Nudge approached the day from behavioural science rather than technology, but his session fitted the same wider theme. His presentation covered concepts including social proof, scarcity, anchoring, curiosity, framing, humour and the Pratfall Effect.

The useful reminder was that platforms and technologies can change rapidly, while the humans using them still respond to many of the same behavioural influences.

The Pratfall Effect was a particularly memorable example. A person or brand that is already seen as competent can sometimes become more likeable by showing an imperfection. Ryanair was discussed as an example of a brand that frequently acknowledges or jokes about its own perceived weaknesses rather than behaving like a perfectly polished corporate entity.

There is an important qualification, though. The competence has to exist first. Self-deprecation can make a competent brand more human, but it cannot turn a poor product into a good one.

The same issue surfaced in Phill’s example involving the band Geese and manufactured virality. Artificial activity can potentially generate early momentum, influence algorithms or create apparent social proof, but none of that guarantees that the underlying product is actually desirable.

My immediate reaction was that if the song is poor, another thousand views simply means another thousand people discovering that they do not like it. Distribution can amplify quality, but it can also amplify mediocrity.

That is why virality and effectiveness need to be separated. A viral event can create enormous exposure, but the more useful questions come afterwards. Do people return? Do they search? Share? Subscribe? Buy? Recommend? Remember?

The principle applies just as strongly to AI-generated content. Having the ability to create more does not automatically mean creating more is a sensible strategy.

In fact, one of the biggest risks of generative AI is that it solves the production problem so effectively that organisations accidentally create an attention problem. If every company can suddenly produce ten times as much content, customers do not magically gain ten times as much attention with which to consume it.

The opportunity may therefore be to use AI to remove low-value production work and reinvest the saved time into higher-value thinking. Research can become better. Interviews can become deeper. Creative ideas can be given more attention. People can spend more time talking to customers, understanding problems and finding the opportunities that do not yet exist in the data.

That feels like a healthier future for marketing than an arms race based on who can publish the most.

The organisers of Social Media Conference Cymru deserve real credit for putting together a programme that explored these issues without turning the day into either AI evangelism or AI panic. The speakers approached the subject from very different perspectives, while the delegates brought thoughtful questions and conversations that made the sessions feel grounded in the reality of day-to-day marketing.

I left the conference more interested in AI than I arrived, but also more convinced that the interesting advantage will not come simply from having access to the tools. Everybody will increasingly have access to similar technology.

The advantage will come from what you know, what you can access, the questions you ask, the quality of your judgement and the originality of the ideas you are willing to pursue.

And that makes Connock’s challenge a useful one for marketers to keep in mind: rather than only using technology to reproduce what has already worked, spend some time looking for the thing the algorithm does not know yet.