Nobody knows anything

Today’s post comprises my musings on the potential impact of AI over the coming years.

The future is unwritten

As I wrote a couple of weeks ago, AI clearly has the potential to be bigger than the internet, or even the Industrial Revolution. 2026 could be a year of significant change, but there are more questions than answers at the moment:

• Massive investment into AI in recent years is positive for GDP growth in the short term, but when will we see a return on this (eg, a boost in productivity)?
• Will the current leading firms in the area (“hyperscalers” such as Microsoft, Alphabet, Amazon, and Meta) be the eventual winners?
• Do current high valuations mean that a bubble is about to pop?
• What will happen to any displaced labour?

I’m particularly interested in:

1. How to monitor for and protect against a potential crash
2. How AI might affect the investment process itself for private investors

But I also look at how AI might impact the economy and, by extension, the markets.

Note that these ideas in many cases come from my recent reading on the topic – I’m not claiming originality but rather trying to collect my thoughts in one place.

Displacement

The key issue at the moment is the extent to which AI will displace human cognitive (white-collar/clerical) labour.

• It seems very plausible that over the next few years, more labour will be displaced than can be absorbed by new markets and services.

Human intelligence is currently scarce and expensive, and this could be about to change, with the rise of agentic AI as the catalyst.

• AI is cheaper, doesn’t sleep, get sick or take holidays, and doesn’t require lossy coordination.

Domains

One reason that there is so much buzz around AI at present is that a key domain where AI has shown progress is software engineering (coding), which is very important to many of those working on AI.

• Indeed, it’s likely that code has been a focus for AI because of the feedback loop by which AI that can code can help to build the next generation of AI.

Many workers in other areas will not have seen this progress up close, and may remain sceptical for now.

My primary focus is investment and economics, and I expect similar progress in this area over the near term.

But the area most at risk in the short- to medium-term is the intermediation/negotiation space.

• For consumers, AI agents could lead to better choices and lower prices.
• But for some of the companies in the middle (software, consulting, financial services, insurance, travel, real estate, payments) this could mean lower profits.

Unevenness

The economic impact will be highly uneven across sectors, regions, and demographic groups (dependent on the regulatory response).

Nor will the AI displacement proceed at the same pace in all areas:

• Physical work will not be affected much by the first wave
• Instead, the focus will be on knowledge-intensive and data-driven industries

But one of the cruellest aspects of the revolution is the targeting of entry-level work.

• Those not already established in an industry might find it harder to become secure as experienced personnel leverage AI to do the work of 10 or a hundred entry-level workers.
• They could be forced into blue-collar jobs and gig economy work (depressing wages in these areas).

Ironically, some of the work which requires the most training and experience to perform well (accountancy, law, project management) will be the first to be affected.

Speed of implementation

The speed at which AI is rolled out through the economy will have a big impact on its key effects.

• Fast implementation could mean that productivity gains can’t keep up with labour displacement.
• Slower implementation would give the economy more of a chance to redeploy the displaced labour and would reduce the potential negative economic impacts of AI.

Many will argue that even though the speed of model progress can’t be currently denied (though some believe that it will “hit the wall” in the near future), the speed of adoption in large corporations is open to debate.

Large companies move slowly, but they also operate in competitive markets, and those who move slowest risk being out-competed by a faster rival or displaced entirely by an upstart.

• A key factor will be whether incumbents resist change (as per Kodak, Blockbuster, Nokia and BlackBerry) or whether they race to adopt AI themselves.
• This positive feedback loop has no natural brake (until everything has been moved over to AI).

A useful indicator might be the long-term bond yield.

• If we are moving “too fast”, then the 10-year yield could start to fall.

Though the AI cycle is atypical in that it does not contain a self-correction mechanism, there are physical requirements inside the loop that could make things go slowly:

• Can we build enough AI hardware (chips, memory, storage, etc.)?
• Can we deliver the power where it needs to be?

And there is also politics to consider:

• Will NIMBYs and Luddites prevent datacentres from being built?
• Will governments choose to slow down AI deployment (despite the risk of being outcompeted by countries that don’t)?

Economic impact

Developed economies rely heavily on their best-paid workers, and many of these will be at risk.

• Without their spending, other companies could be forced to scale back, depressing the economy.

We also need to think about mortgages.

• What happens to real estate if a significant proportion of borrowers can’t make their repayments?

And welfare payments from the government might rise just as tax receipts fall, leading to increased budget deficits.

• In theory, taxes could be raised on AI agents, but this could be tricky in practice and might lead to a race to the bottom between national economies.

Sectors

Things look bad for:

• Anything with lots of clerical and analytical  workers
• Intermediation services (eg, insurance)
• Repeat (inertia) subscriptions and orders
• Software (witness the recent “Saaspocalypse”)
• Property
• Payments (eg, credit cards)
• India and the rupee (because of its large IT services sector – Indian coders are cheaper than Western coders, but more expensive than AI)
• Companies with “moats” which were previously deemed to be impenetrable

Stability should be possible in:

• Construction (but who will buy the new houses?)
• Healthcare
• The public sector (which is bound to transition at half the pace of the private sector)
• Resources
• Utilities
• Possibly some aspects of financials
• Companies with tangible rather than intangible assets

There are opportunities for:

• Crypto, if it can be set up properly as a cheap payment system
• Countries like Taiwan and Korea, which provide a lot of AI picks and shovels
• Autonomous vehicles, if the kinks can be worked out in time

Whether AI favours active or passive investing depends on how easy it will be to identify the winning and losing companies.

Investment Impacts

There are a variety of conceivable changes to the investment process for a private investor (PI), though I am not aware of any of these happening as yet.

1. At the highest level, costs should come down and/or more sophisticated approaches should be available at today’s cost levels. This may make some complicated strategies more accessible to PIs. 
2. There is potential for tools used by institutional investors to filter down to PIs. Whether things like sentiment analysis or the use of alternative date would be of much use to most PIs is another question.
3. There is obviously scope for increased personalisation of investment solutions, but whether this will be true personalisation or a repeat of the high-touch marketing process backed by a few standard portfolios that we saw with the first generation of robo-advisors remains to be seen.
4. Conversational interfaces to make portfolio changes easier are likely.
5. Tax reporting (for GIAs) and tax optimisation of brokerage accounts (more important in the US than here in the UK) are likely.
6. Something to watch out for here in the UK is over-regulation. The FCA often takes a safety-first approach, “protecting“ UK investors from innovation until it’s too late to take advantage.
7. At the same time, it’s clear that AI investing tools will resurface the “black box” issue of limited process transparency. This. might in turn slow mainstream adoption.
8. There’s also a risk of overfitting, as with algorithmic and systematic investing in general, and the issue of transferring strategies between market regimes (and of how to identify these regimes). 
9. The concentration of providers could be an issue and could increase the impact of any cybersecurity threats.
10. Herding of trading behaviours from the widespread adoption of similar AI models could increase market volatility and instability.
11. The importance that can be attached to qualitative/narrative input  (stories about stocks) should decline in an era where anyone can generate a good story. This should mean that quantitative input (numbers) should be relatively more important.

Conclusions

I doubt that reading this article has made things much clearer for you – writing it certainly didn’t firm up too much on my end.

• I still believe that we don’t know too much about how far this revolution will go, how unevenly this version of the future will be spread, and crucially, how quickly things will happen.

But I think that the next couple of years might make the last couple of years seem like smooth sailing, and I want to be ready.

So as I mentioned a couple of weeks ago, my priority for 2026 is getting my affairs in order, in anticipation of a storm that might hit in 2027, 2028 or perhaps never.

I have a house to sell, but beyond that, I will be focusing on:

• My outgoings
• My asset allocation
• Crisis alpha
• Leverage (that can be easily dialled up and down)
• TAA
• And Dashboards to see what’s happening and to predict what’s around the corner.

That’s it for today.

• Until next time.