Most of us love a bit of fast food. It’s convenient, it solves a problem quickly, and sometimes is just what you fancy (especially on a late night train back up north from client meetings!).
I’d never suggest banning it, and certainly wouldn’t judge anyone for eating it, but I’d also probably not want to live on it. Recently, I’ve started wondering whether AI is becoming the fast food of thinking.
I say that as someone who is enormously positive about AI. I use it, the team at Marmalade Marketing uses it, and I spend a lot of time helping clients work out where it can genuinely improve their businesses. I studied artificial intelligence at The University of Oxford because I wanted to understand what was happening properly, rather than form an opinion from whatever happened to be trending on LinkedIn that week.
The more I learn about it, the more exciting I find it. The more I use it, the more interested I become in what all that convenience could eventually do to us as professionals.
AI removes friction in business processes, without doubt. A question that might have sent you down a Google rabbit hole can be answered in seconds. You can get AI to explain something that’s puzzled you for years like you’re a child. If you process information differently, you can ask as many follow up questions as you need without worrying that someone will start impatiently tapping their foot waiting for the penny to drop.
For people who are isolated through circumstance, technology can open a door to a world that might otherwise feel very small. The UK AI Security Institute found that a third of UK adults surveyed had used AI for emotional support or social interaction during the previous year. For some people, there will be enormous value in having something available when human support is difficult to reach.
All of that makes me optimistic, but it also makes me think about chips. Stay with me!
Fast food works because it removes effort between wanting something and getting it. AI pretty much does a similar thing with thought.
There’s something quite enticing about never having to sit with the annoying bit of a problem for very long. The blank page disappears, in its place an answer that’s arrived in seconds. Even a fairly woolly question can come back with something polished enough to make you feel like you have made progress.
We’re already starting to see the cultural result in the mountains of what people now call AI slop. Content appears because it can be produced, rather than because anybody had anything particularly interesting to say. I’m less concerned with this making LinkedIn more boring (we’ll survive), but I’m more interested in what happens when the same behaviour is applied to work that requires judgment.
Last year Deloitte agreed to refund part of a $440,000 Australian government contract after errors were found in a report that had used generative AI, including nonexistent references and an incorrect court reference. The conclusions of the report were not withdrawn, but the episode became a very public example of what can happen when convincing output receives more trust than it deserves.
We’ve talked about this at length in the office; false confidence is fast becoming one of the more interesting problems in AI adoption. AI is very good at sounding as though it knows, and humans are quite good at believing something that sounds as though it knows.
Put those together inside a business chasing efficiency and you can see why governance suddenly becomes much more important.
I remember the early days of social media feeling incredibly exciting. Ordinary people could publish without needing permission from a newspaper or broadcaster. Communities formed around interests that would previously have been difficult to find locally, and small businesses suddenly had a way to speak directly to customers without needing enormous media budgets.
I built a career in marketing while a lot of that change was happening, so I am hardly going to pretend it was all a terrible mistake.
However, once businesses realised our attention had value, platforms became incredibly sophisticated at understanding how to keep it.
We now have infinite feeds because stopping creates a moment where somebody might decide they have had enough. Recommendations become more personal because relevance keeps us interested. The experience gets smoother because friction gives us opportunities to leave.
The UK Government has reached the point where it is deliberately putting some of that friction back. In its 2026 consultation on young people and online life, 81% of respondents considered infinite scroll particularly persuasive and personalised recommendations followed closely at 78%. New protections will switch autoplay and personalised feeds off by default for 16 and 17 year olds, alongside overnight restrictions.
That’s quite an extraordinary turn when you think about it. We spent years getting incredibly good at removing the moments where people might stop using technology, and now we’re designing some of those stopping points back in.
The wider social picture is complicated and I would never claim that social media alone created a loneliness problem. There are far too many factors involved for that to be credible.
But it’s still hard to ignore the irony. The Government's new Social Connection Framework describes young people as the most connected and yet the most isolated generation in history, with young people five times more likely to experience chronic loneliness than older people.
Connection became abundant, but it didn’t automatically make us feel more connected.
I wonder whether AI will eventually give us a similar lesson about knowledge: answers are becoming abundant, but I’m less certain that having an answer instantly will always make us better at knowing what to do with it.
This is probably where the marketer in me becomes impossible to switch off. Whenever behaviour changes at scale, somebody eventually works out how to build a business around it.
Recruitment has lived through its own version of this. Reed.co.uk launched in 1995 as the first recruitment website offered by a UK recruitment agency, and became the destination for people looking for work. As job seeking moved online, access to that audience naturally became commercially valuable. Reed today operates one of the UK's biggest jobs and careers marketplaces.
There’s nothing sinister about that; Reed built something useful and businesses paid to participate in an ecosystem where their audience was already spending time.
Social media followed a similar economic logic with attention.
AI could take that model somewhere even more interesting because the information we give it is often much closer to intent. People use AI while they are trying to decide something. They might be researching a purchase or thinking about a new job. A business leader might be trying to understand which software will solve a problem.
It’s all context that’s incredibly valuable to marketers.
ChatGPT Ads launched in the UK in August, and by the end of that month OpenAI said the advertising business had reached a $1 billion annualised revenue run rate. OpenAI says advertising remains separate from ChatGPT's answers and does not influence them, but the speed of commercialisation still tells us something important about where the market is heading.
OpenAI's own advertising proposition talks about reaching people while they are exploring options and making decisions, because conversations provide richer context than traditional keyword search.
For years, we have tried to infer what somebody wants from a search query, a website visit or a click, and AI can now potentially understand the conversation around the decision itself.
Social media taught businesses how valuable attention could be, but AI may teach us how valuable intention is.
I find the conversation about whether people should or should not use AI slightly tedious because the answer is already obvious.
They are going to use it, and I am too.
It’s much more useful to consider whether we are developing the judgement to use it well at the same speed as we are developing the habit of using it.
There’s a version of AI adoption that makes somebody significantly better at their job because the repetitive work becomes easier and they have more capacity for the work that needs their experience. Still, there’s another version where somebody slowly becomes unable to start without asking the machine first.
If a business tells me AI adoption is up, I want to know what people are actually trusting it with. I want to know whether expertise is being strengthened or quietly replaced by a very convincing average of what the internet already knows.
Those questions become even more important as AI starts touching decisions that affect other people.
I’ve been writing a lot about AI in recruitment and employment, where regulators and courts are increasingly grappling with how much transparency somebody deserves when an automated system influences a consequential decision about them.
Once AI moves beyond helping us draft the thing and starts helping decide ‘the thing’, the standard has to change.
I don’t think AI is inherently bad for us. That would be a strangely simplistic conclusion after everything I have learned about it and everything I have seen it make possible.
I do however think we’re at the stage where convenience is so intoxicating that we need to pay attention to what becomes habitual.
Fast food has never been a problem because somebody had a burger, but when convenience becomes the whole diet and nobody remembers what nourishment feels like? That’s a problem.
AI can make expertise more accessible. It can remove work that consumes time without adding much value. It can give people access to information and support they might never otherwise have had. I want all of that. But I also want people to remain curious enough to question the answer, knowledgeable enough to spot when something seems wrong and confident enough to occasionally close the chatbot and think for themselves.
Perhaps the lesson from social media is that we should have these conversations while a technology is becoming embedded, rather than waiting until it is so embedded that we have to legislate the friction back in.
I have no intention of putting ChatGPT on a diet. But I am going to pay much more attention to what we are feeding our brains. Because there is nothing wrong with the occasional fast food treat; I just don’t want us to forget how to cook.
If your business is moving quickly with AI and you’re trying to work out where it genuinely adds value, where people still need to use their judgement and what sensible guardrails should look like, that’s a conversation we’re having a lot at Marmalade Marketing. If you want to compare notes, drop me a message.