On Friday the Nasdaq fell by almost three per cent, its worst day since early June. Intel lost more than a quarter of its value in a single session. Alphabet had already suffered its worst day in over twenty years, and Amazon’s spending on AI had pushed its cash outflow to a record.
Next week Apple reports on Monday. AMD reports on Tuesday, the same day America votes in its midterm elections.
I ended last week’s post wondering what it would take for us to pay for presence rather than for a partial record of past performance. Looking at what people are doing now and what they need, and less at what they once did that disappointed or excited us for reasons we did not understand at the time.
This week has given the question an edge. Perhaps it was Halloween…..
How much of Friday was a change in the story about whether AI will pay back, how much was the midterms and what they might do to the balance of American power, and how much was speculation for its own sake, nobody can really say.
That uncertainty is what has been on my mind this week.
Economics always lives in the space between the data about what has happened and the stories about what we expect to happen. Neither is new: data describe the past, stories describe the future. The market is an exercise in judgement that sits between them.
What seems to have changed is how far our notion of value leans on the story. I have come to think of the gap as a rubber band, stretched between the evidence of the past and the stories people tell about the future, increasingly, it seems, to justify hubristic decisions already made in a moment of excitement. The hundreds of billions committed to data centres are decisions of that kind. On Friday the evidence, in the form of free cash flow, may have signalled the beginning of those stories being called to account.
Or perhaps we will just be given bigger, better, shinier stories.
Stretch a rubber band far enough and one of three things happens. It breaks, snaps back to where it started, or, held at full stretch for long enough, it loses its elasticity and forgets it was ever a rubber band. The tension that gave it power disappears.
The third is the one that worries me. A market that stays stretched long enough forgets what normal looked like, and has nothing left to return to. We end up somewhere closer to Through the Looking-Glass than to rational economics, running as fast as we can to stay in the same place, without knowing why.
I think there is another way to see the same thing, which explains why the band stays stretched. Feedback. Hold a microphone too close to the speaker it feeds and you get the howl. It starts when the gain round the loop goes above one, so that each pass comes back louder than it went out. Speaking louder makes it worse. The cure is to turn the gain down or move the microphone.
Keynes described the financial version in 1936. In the General Theory he compared professional investment to a newspaper competition in which the prize goes to whoever picks the faces the other entrants will pick. The skill he described was judging what others will think others will think: a loop listening to itself.
Whether in markets or in politics, our stories now run through more gain than they used to. Attention-hungry media amplified them before. Now AI amplifies them too, producing new stories in response to the ones it hears, at speeds no human newsroom can match. Early research on models trained on their own output finds that the rare and the original drop away first. What remains is the average, amplified: more commentary, fewer new observations.
Howling mediocrity. It makes it harder to see what is actually happening, and harder still to reflect quietly when every fibre of us is reacting to the noise.
For most of us, jobs are where this matters most.
On Monday Microsoft offered voluntary exit packages to around 60,000 staff, roughly a quarter of its workforce. People working on its data centres, sales and marketing were among those excluded, (keeping the gain at maximum).
It follows a voluntary retirement scheme in the spring and around 4,800 job cuts in July. The week before, Barclays said it would shed about 5,000 jobs by the end of the year, roughly one in twenty of its staff, many of them in support functions and many of them developers. It announced the cuts alongside a twelve per cent rise in profit and a £500 million share buyback. Its chief executive called AI “the icing on the cake, not the cake”.
Microsoft says it expects fewer than one in twenty of those eligible to accept, which would be nearer three thousand, and that AI is not driving the programme. The gap between those two numbers is the howl in miniature.
If I thought these cuts reflected a settled reality about AI, I would be more relaxed: we would have a problem we could get to grips with. I am not sure we do. The howl is that AI is replacing jobs, yet the evidence so far says something narrower. The effect on overall employment appears muted, and two thirds of firms in McKinsey’s latest survey report little or no AI-related change in headcount.
Where the effect is real, it falls on the young. Employment of people in their early twenties in the most exposed occupations is running well below where it would otherwise be, and if AI does not fully live up to the hype, our organisations will find themselves running on empty when it comes to new talent when the howl allows them to hear themselves think. AI is reshaping work and removing the first rung of the ladder, which is serious in its own right. One-for-one replacement of people by software is a different claim, and the evidence does not yet support it. The cuts are being made on the strength of the story, ahead of the evidence giving us the howl we all sense.
The first rung matters. It was where people learned the judgement that organisations now expect to be provided by technology. Remove it, and in ten years there will be nobody senior who learned the work by doing it.
Watching it is a little like watching someone go slowly mad. Established organisations, built on the stories, data and cultures of the past, do not know what to do as the world changes, deafened as they seem by the howl.
The next few weeks may move the microphone a little further from the speaker. They also may not, when the money depends on the howl.
Either way, trusting our futures to organisations in that state does not seem a sensible thing to do.
Peter Drucker argued that the most important work of an executive is to identify “the future that has already happened”: the changes already under way whose effects have not yet arrived. You find them by looking out of the window at what is visible but not yet seen, by being present.
Think of a ward sister who senses a change in a patient before the data does, or a teacher who notices which child has gone quiet. Neither is forecasting. They know what is likely to happen next because they are there, paying attention, inside relationships that let them sense what no algorithm would.
That is the practical answer to last week’s question. To see where the future is going, we have to be present where we are, in the space between the burden of past data and the myth of future stories. Presence means being involved, and being involved means relationship: the people we work with, and the things we notice because we are there to notice them, because we care about them. None of that shows up on a dashboard. It is the raw material of judgement, put to work for something that matters more than the next retrospective report.
This is what we would be paying for if we paid for presence. Performance is what we witness. Presence is a relationship with the thing itself, taking its bearings from the room instead of the loop. It is how a person stands far enough from the speaker to hear what is going on, to see the performative for the shabby proxy of what genuine performance might yield.
We cannot stop the howl by shouting louder.
Perhaps, though, we can help move the microphone away in the way we use social media, LinkedIn and the rest. Be thoughtful before we like, repost, or shout back.
Turn down the gain, reduce the howl, and find the space to think clearly.
‘Like watching someone go slowly mad’ is an apt way to describe the behaviour of companies in recent years. They have succumbed to the siren call of the story, whilst telling themselves they were ‘following the data’.
AI is not removing entry-level jobs, CEOs are. It is a mistake that will come back to hurt the companies but the CEOs will be long gone, spending more time with their money.
What we saw on Friday was the moment when the little boy speaks in The King’s New Clothes. Now, there’s a really useful story…