I’m drinking the AI Kool-Aid. Sorry.

AI is humankind's last invention. It's much more likely that AI is going to kill us all than help us. And until recently, it could not reliably tell you how many r's are in “strawberry.”

But....

A few days ago I realised how far gone I am. I said that if AI disappeared tomorrow, I would quit my job at Change Strategies and try to make a living building furniture instead. I am not a carpenter, so this would probably be a terrible career move. But if I suddenly had to go back to doing all the research, calculations, analysis, documents and production work manually, I would rather do something that is actually manual. At least at the end of the day there would be a table standing in front of me.

A few days later, I said I was drinking the AI Kool-Aid. The expression comes from Jonestown, where more than 900 people died in 1978 after a poisoned drink was distributed to members of the Peoples Temple; many of the victims were children, and not all of the deaths were voluntary. The phrase survived as shorthand for following a belief somewhere dangerous. That is why it works for me. I know this particular drink may be poisonous, and I am still drinking it.

AI is bad because it is stupid. AI is bad because it is genius.

When I first started building things with GPT-3.5, I showed people what it could do because I thought it was extraordinary. Their reaction was often to show me what it could not do. It hallucinated. It misunderstood simple questions. It confidently produced nonsense. For a while, one of the favourite demonstrations of AI stupidity was asking a model how many r's there are in “strawberry” and watching it get the answer wrong.

I had the opposite reaction. Of course it made ridiculous mistakes. What fascinated me was that the thing existed at all. A machine could suddenly write code, explain complicated subjects, work with language and help me solve problems I would previously have needed another person, a piece of specialist software or several hours to solve. I did not need GPT-3.5 to be good at everything to see what would happen if it kept becoming good at more things.

A few years later, AI has somehow managed to become both embarrassingly stupid and existentially dangerous. It creates generic content nobody wants to read, but it is also going to destroy creative industries. It cannot reason properly, but it is going to take everyone's jobs. It produces AI slop, but it may also create a technological singularity and become more intelligent than we are. It consumes too much electricity answering stupid questions, but it might transform science, medicine, education and almost every knowledge industry.

Apparently, AI is bad when it cannot do something and bad when it becomes frighteningly good at it.

I find the environmental version of this particularly strange. AI absolutely uses energy and water, and rapidly growing data-centre demand is a real problem. But data centres did not appear with ChatGPT. We have spent years building enormous computing infrastructure to store social-media photos, optimise advertising and decide which video should keep us scrolling for another twenty minutes. I do not remember the environmental cost of computation being nearly as culturally offensive when the result was a better Instagram feed. Now some of that same infrastructure is helping people write software, analyse businesses, learn languages or understand complicated documents, and suddenly “computers use electricity” is treated as an argument against the usefulness of the technology.

AI slop is similar. There is a lot of it, and much of it is terrible. But social media was hardly a pristine collection of carefully considered human thought before generative AI arrived. We already had clickbait, misinformation, rage bait, harassment and industrial quantities of content nobody needed. AI has made producing things much cheaper. Predictably, that means more good things and more bad things. The technology amplifies capability. What humans do with that capability remains our problem.

And that is where I think the debate becomes more useful.

Nobody gets to decide whether AI happens

Around the time I started experimenting with GPT-3.5, I was learning about game theory. I am not claiming that global AI development fits neatly into one textbook game, but it gave me a useful way to think about incentives.

If a technology can create an enormous economic, scientific or strategic advantage, restraint only works when you believe everyone else will show the same restraint. If one company stops and another continues, the second company may win. If one country stops and another continues, the stakes become much larger. Expand that across governments, research labs, corporations, open-source developers, entrepreneurs and millions of individuals with completely different laws, ethics and circumstances, and I find the idea of collectively deciding to uninvent AI very difficult to take seriously: Somebody will keep going.

A professional in Germany can decide that using AI to write something is intellectually impure. That is their choice. Somewhere else, another person is looking at exactly the same technology and seeing a chance to compete with people who had better schools, more money, stronger institutions and access to specialists they could never afford. Others see scientific progress, military advantage, money or simply the irresistible possibility of discovering what comes next.You are not going to get all of those people to stop.

This is the part I think gets confused most often. Whether I personally believe AI is worth the risk is almost irrelevant. Nobody appointed me to make that decision, and nobody appointed you either. Development is happening because millions of actors have reasons to continue it.

That does not make regulation useless. It does not make safety work useless. We should obviously try to prevent harm. But “I don't want AI to happen” is not a strategy for living in a world in which AI is already happening.

For a business, that distinction matters.

I don't want my old job back

AI is the most useful tool I have ever had. That is not a prediction about what it might become. It is a description of how I work today.

It removes a huge amount of work I never enjoyed in the first place: bureaucracy, paperwork, accounting preparation, documentation, tedious research, technical troubleshooting and the endless reshaping of information from one format into another. More importantly, it has expanded the work I can realistically do. I can build simple applications without being a developer. I can analyse quantities of information that would have been unrealistic for me to process manually. I can create voice, video and advertising assets without assembling a production team. I can enter an unfamiliar field and understand enough of it quickly to know where I actually need an expert.

That is why the furniture comment was not really a joke. AI has not replaced my job. It has changed the ratio between the parts of my job that require judgment and the parts that simply consume time. If that ratio suddenly went back to what it was before, I do not think I would want the old job anymore.

I have no emotional attachment to spending three hours on something a machine can do well in three minutes. Doing work inefficiently does not make it more human. If AI can take a task from me, I can use those three hours to think about a harder problem, work with another person, take better care of the business or go surfing.I know both versions of my working life now. I know which one I prefer.

The same technology looks very different from a pueblo

The part of the AI conversation I find most exciting becomes much clearer when you leave the wealthiest professional bubbles.

I spend a large part of the year in Central America. In the major cities, businesses have access to good designers, lawyers, accountants, doctors and developers just as they do elsewhere. But move into a smaller town, a modest business or a family without much money, and access to expertise changes very quickly.That is where AI becomes something other than a productivity tool.

Someone in a pueblo receives a government form full of language they do not understand. The person behind the desk in the city deals with these documents every day; the person holding the form might encounter one twice in their life. That is a huge knowledge imbalance. Now they can photograph the form, ask what it means, ask again in simpler language and arrive at the government office knowing what information is missing and which questions to ask.

The same applies to understanding a contract before speaking to a lawyer, making sense of medical information before seeing a doctor, learning how to repair something, getting help with schoolwork, writing a business plan or communicating professionally in a language you do not speak perfectly. AI is not a substitute for expert judgment when expert judgment is actually required. What it changes is how helpless you need to be before you reach the expert.

I grew up in Germany, where access to qualified professionals and functioning institutions was normal enough that I barely thought about it. Today, if I have a straightforward legal question about a document, I often use AI first. I can understand the document, identify the actual uncertainty and then call a lawyer if I need legal judgment rather than paying someone simply to translate specialist language into normal language.Now imagine you never had easy access to the lawyer at all.

This is why the World Bank's argument about AI and developing economies matters to me. Cheap access to knowledge and expertise creates the possibility of leapfrogging parts of the infrastructure that wealthier countries took generations to build. Education, healthcare, agriculture, government services and small businesses can all benefit from the same basic change: one person with a phone can suddenly access capabilities that previously required money, geography, formal education or connections.

For someone who already has everything, AI may save an hour.For someone who had very little access to expertise, it can change what is possible.Try taking that away again.

The drink may still be poisonous

None of this makes me think AI is safe. It is almost the opposite. The reason AI can create such extraordinary benefits is the same reason it can become extraordinarily dangerous: the capability keeps increasing.

I do not know what happens if we eventually create something vastly more intelligent than ourselves. I do not have a convincing argument for why such an intelligence would automatically value the continued existence of billions of humans.Maybe AI has mercy.That is not a strategy. It is hope.

But hope is not the reason the technology will continue developing, just as fear is not enough to stop it. The development continues because the incentives behind it are bigger than any one person's opinion.

So I look at the situation we actually have: For most of my life, we have known about many of the largest threats facing us. Climate change. Nuclear weapons. War. Pandemics. Fragile financial and social systems. Some of these problems have been obvious for thirty or forty years, and our ability to discuss them has consistently been much greater than our ability to resolve them.

AI is the single best chance I can see of changing that equation because it changes the amount of intellectual capability available to work on the problems. It can help researchers test more ideas, scientists work through more information, people learn faster, individuals access expertise and small teams attempt things that previously required institutions. None of that guarantees a solution. But for the first time, the limiting factor in some of these problems may move significantly.

That is why I am optimistic despite being genuinely frightened by the technology.AI may create problems worse than the ones we hope it will solve. But it also gives us a chance to attack old problems with capabilities we have never had before.I would rather live through that than watch the old system fail slowly.

The Change Strategies position

At Change Strategies, our position is not that AI is good, safe or ready to run your company while you go to the beach. Our position is simpler: reality comes first.

Marketing strategy has to work with the product you actually have, the customers you actually have and the world you actually operate in, not the version you wish existed. AI is now part of that world. Your employees use it. Your customers use it. Your competitors use it. We use it.

So the interesting business question is not whether AI is good or bad. It is what has become possible because it exists. Which costs have collapsed? Which bottlenecks have disappeared? What can a small team now do that required a much larger one a few years ago? Which assumptions in your strategy stopped being true while everybody was arguing about the technology?

There will also be places where AI is unreliable, inappropriate or simply worse than a human. Then don't use it there. The point is not to automate everything. The point is to keep looking honestly at what the technology can do now and adapt when that changes.

If AI disappeared tomorrow, I would probably be in my workshop trying to work out how to build a chair people were willing to pay for. Fortunately, I don't think that is going to happen.The Kool-Aid may be poisonous. I understand why people are afraid to drink it. But refusing the cup does not put the technology back in the bottle.

So yes. I'm drinking the AI Kool-Aid. Sorry.

Sources

Alternative Considerations of Jonestown & Peoples Temple, San Diego State University, Drinking the Kool-Aid: An Introduction and related historical material on Jonestown.

OpenAI, Introducing ChatGPT and Introducing ChatGPT and Whisper APIs.

International Energy Agency, Energy and AI.

World Bank, World Development Report 2026: The Promise of Artificial Intelligence.