I’m sceptical about AI. But this is precisely why I am optimistic about AI.

The hype has earned every eye-roll it gets. The evidence has earned a good second look.

There is a particular silence I have learned to recognise. It arrives about four slides into a pitch, just after the word “AI” has appeared for the third time. It also appears when someone promises to reinvent an industry none of us has finished understanding.

In that silence, somebody at the back of the room folds their arms. I know that person well. For a long stretch of the last few years, that person was me.

So let me say the unfashionable thing first, to earn the right to say the hopeful thing second.

Move beyond the hype to the usefulness

Most of what you have been told about artificial intelligence is marketing. The breath-taking launches, the valuations that read like typos, the founders who talk about the future the way other people talk about religion. None of it is evidence, and you were right to distrust it. If you have stopped listening, I understand why. I nearly did, too.

What brought me back was not a better sales deck. It was a footnote. Earlier this year, the Federal Reserve Bank of San Francisco published an Economic Letter, the sort of document written in the deliberately unexcited prose of people whose job is to be unimpressed.

The letter reported that spending on information-processing equipment, software, and data centres accounted for roughly a third of all business investment in the United States in late 2025. That was the highest share since 1947. You cannot dismiss a number like that as a vibe. It is the plumbing of an economy quietly being re-laid.

Investing heavily in AI

The same researchers point out that almost all of this growth came from a handful of the largest firms, Amazon, Alphabet, Microsoft and Meta among them, and that if the optimistic forecasts do not hold, the cost of that bet will fall on a very small number of shoulders.

They note that such concentration could itself raise prices and slow the very adoption everyone is counting on. That is the tone I have come to trust. Optimism you can audit. Enthusiasm that arrives with its own list of ways it could be wrong.

The clearest articulation of that stance I have read comes from Miles Brundage, in an Oxford essay titled “Scaling Up Humanity,” which I keep returning to. His argument is what he calls conditional optimism, and the condition is not a footnote. It is the whole sentence.

The heavy lifting

If and only if we do the unglamorous work of solving the technical and governance problems, the upside is genuinely large. The reason is uncomfortable and worth confronting: the very properties that make these systems capable of doing enormous good, their competence and their scalability, are the same properties that make them capable of enormous harm.

The essay does not look away from the harm. It names automated hacking, misuse, and the pressure on responsible builders to cut corners when competitors do. And then it makes the case for the good anyway. And the good list is heartening.

The researcher says AI will help deliver faster routes to cheap, clean energy and solve health problems. AI will also help to build public services that cannot be quietly corrupted. This is not naive thinking. It is the opposite of naive. It is a pragmatic hope that has read the risk assessment.

AI is a game changer

As an African entrepreneur who loves his continent, this matters. Affordable clean energy, good health systems, and incorruptible public services are game changers.

When the World Economic Forum, working with PwC, went and asked entry-level workers around the world how they actually feel about AI, the answer confounded the doom narrative.

Across all regions studied, young workers reported being more curious (47%) and more excited (38%). Only 29% were worried about AI. The strongest excitement anywhere was not in Silicon Valley but in India, Malaysia and Türkiye. The people standing closest to the disruption, the ones with the most to lose, turn out to be the least cynical among us. That should give the folded-arms crowd some pause for thought.

Job security is still an issue

I will not pretend the same survey did not trouble me, because it did, and the trouble is the point. Nearly one in three of those same young workers is anxious. Only just over half feel secure in their jobs.

And there is a gap I cannot stop thinking about: business leaders and their newest employees are looking at the identical technology and reading it completely differently, with roughly as many executives expecting AI to cut entry-level roles as are expecting it to create them. The report calls this a challenge that is not only technological but perceptual.

I would like to put it more plainly. What we have here is a trust problem. There is an older cohort, the workers in their fifties, changing careers or re-entering the workforce, who have barely touched these tools. We need to change this because they risk being quietly left behind.

Where optimism and scepticism shake hands

This is the part where my scepticism and my optimism finally stop arguing and shake hands. Because if you take the Fed’s warning about concentration, and Brundage’s insistence that the benefits are conditional, and the Forum’s evidence that the anxiety is really about trust and being left behind, they all point to the same conclusion.

The technology is not the variable that decides how this goes. We are. Every genuine upside on offer is contingent on choices that have nothing to do with model size and everything to do with whether we keep human beings at the centre of AI progress.

This is a conviction I have held long before I read any of these papers, and it is the one they have made harder, not easier, to dismiss: technology has to stay in service to people, or it is not progress; it is just acceleration. In fact, this is the very subject of a talk I give frequently. It is called Optimism in the age of AI. Here is the link: https://www.youtube.com/watch?v=6_jX6VeQQ0E&t=942s

Scaling the AI of humanity

So when the Oxford essay talks about scaling up humanity, I do not read it as a slogan. I read it as an instruction. The compute will scale on its own; the money has already decided that.

What will not scale automatically is judgement, empathy, care, and the patience to bring the fifty-five-year-old and the nervous graduate along. We cannot afford for anyone to be left behind.

Those are the things a hype cycle never budgets for, and they are the entire difference between a future worth having and a very expensive one that only works for four companies.

Scepticism that is useful

Which brings me back to the person at the back of the room with the folded arms. I used to think that posture was the opposite of optimism. I have come to believe it is a form of it.

You only bother to roll your eyes at a broken promise if some part of you still expects better. So keep the scepticism. Keep demanding the footnotes, the caveats, the evidence that has been stress-tested by people paid to be unimpressed. That habit is not the enemy of hope about AI.

Hype is never worth trusting and doesn’t build trust or hope. What builds a hope worth trusting is the hard work of ensuring that AI lives in service of humanity, and that we increasingly see more evidence of this.