The Big Question - AI: Where will disruption emerge?

Describe where companies are in terms of AI’s adoption and impact?

Mick Dillon: So AI is rapidly transforming many, many different parts of business. We can see it in advertising, software coding, customer service levels at different companies are seeing tremendous impacts.

If you think about coming out of Covid, there was a real war for talent, and you couldn't hire software engineers for love or money. But if you look back over the last three years, big companies like Meta or Google or Microsoft or Amazon, they've seen their overall headcount over the last 2 or 3 years, only up by about 3%.

But the revenues at all four of those companies are up between 25% and 40%, because software coding in particular is seeing huge gains from harnessing the power of AI.

One other area that we're seeing huge gains is in customer service. We're seeing more calls answered quicker, faster, with better response rates and more accurate response rates. In other words, customers are happier and getting served better, faster, quicker and yet the cost to do this are going down because we need less people and we're getting more accurate answers.

So there's a lot of areas in business where we're seeing really, really big gains from technology right now.

Danielle Menichella: We believe AI will be one of the defining productivity shifts of our lifetimes, and companies are no longer debating whether AI matters. Most have moved beyond “what is it?” and are now experimenting, investing and proving return on investment.

Companies know failure to embrace AI is an existential risk, but adoption is still broad and shallow. Coding assistants are increasingly common among engineers, but usage across broader workers remains very low relative to the roughly 1 billion global knowledge workers who could ultimately benefit from agents. We estimate that penetration to be 1% to 2%.

That means AI's adoption and impact is still early. Many companies are adding AI tools, but few have redesigned their processes around AI native ways of working and that matters because the market can underestimate how early we are.

That gap is why we believe compute, memory, and storage demand can grow structurally over the next decade, even if the path is volatile.

Tye Bousada: We are moving from the experimentation phase to the execution phase. In 2023 and 2024 companies were largely running small pilots and building wrappers around large language models.

Now, the focus has shifted to return on investment and integration, with the goal of driving efficiency. Big technology shifts have been always about driving efficiency, and the impact isn't uniform.

We're seeing a two-speed economy and adoption. Front runners are using AI to re-engineer their core workflows. Think AI-augmented software engineering or AI managing the workflow around commercial insurance brokerage post the sale of a policy. This is where the real margin expansion opportunities are.

Meanwhile, laggards are still stuck using AI for basic automation tasks like writing emails. The real story isn't just using AI. It's the structural shift towards workflows where AI doesn't just suggest actions but executes them and this is where we see the next big jump in productivity coming from.

Is the picture becoming clearer as to which industries may win or lose?

Mick Dillon: I think the picture is becoming clearer as to which industries might win or lose and there's an even bigger issue behind all of this. The question is, is AI a technology risk? And any risk is an opportunity as well? Or is it a business model risk?

The reason we frame it like that is because if it's a technology risk, then the race is on between incumbents trying to harness the power of AI to deliver new services and products for their customers before some AI start-up in a garage disrupts their business and steals their customer. That's a technology risk.

A bigger risk is if this is a business model risk, and suddenly the way that you serve your customer is changing because the business model, as an example, in software, changes from being a software as a service (SaaS) subscription business model to an advertising based business model, where it's free for the customer.

Danielle Menichella: So the big changes are not necessarily from the technology, but what they can do to this structure of companies and business models.

AI is a general-purpose technology so I'd be cautious about declaring entire industries winners or losers. Any company can either be disrupted or become the disruptor depending on how they use AI.

The clearest winners today are the enablers of AI like AI infrastructure, semiconductors, cloud platforms, power, memory, and storage. Longer term, it's unclear how economics work out in things like inference and tokens and training.

The most vulnerable areas are routine, undifferentiated knowledge businesses like point solution SaaS (software as a service), basic content creation, low end outsourcing. Because AI can compress their value proposition and make it easier to replicate.

That doesn't mean those businesses are dead, but harder to analyse. Longer term, the biggest opportunities may be from companies that use AI to change their economics. Like lower margin heavy workflow sectors could use AI to expand margins and improve productivity, as long as they don't compete away those savings.

Tye Bousada: Shorter term, picture is certainly sharper, but there's still a lot of fog if you look out beyond a year or two. The clear winners of the moment are the enablers the silicon and power infrastructure providers and the data owners. Companies with proprietary clean data sets that can't be scraped off of the open web.

The market can see this, and we've seen big price movements in enabler share prices recently. The losers, though, aren't necessarily entire industries, but rather commodity middlemen and business models built on simple information arbitrage or basic content synthesis is under threat as well.

Another category of potential losers are those who treat AI as a line item expense, rather than a core strategic pivot.

Where the opportunity seems to be are the businesses that are perceived as having models that are under threat from AI, but in fact will benefit from them So an example of that would be S&P Global, that has a proprietary data set that AI can't scrape off the internet or brand power that insulates the business from competitive threats.

Do you think the market is getting anything wrong about AI?

Mick Dillon: A prime example of the market struggling to discern within AI is Google or the parent company Alphabet. Since the launch of ChatGPT back in November 2022, Google has gone from AI loser to winner to loser and back to winner again over that three-year period.

If you look at Google today, they've got 13 applications with a billion users powered by AI. They've got the leading language model in Gemini, they've got their own cloud infrastructure, and they've got their own semiconductors that they can use to drive all of that infrastructure.

So, they look like a real winner. But the market has been oscillating back and forth over the last three years.

Danielle Menichella: I think the market’s missing a few things. First, by treating it as AI infrastructure versus all other growth industries. As a result, the perceived AI enablers have seen their stocks go up at the expense of other traditional growth industries like software, internet and other growth businesses that may still prove that AI is a positive.

We think there is still value to flow to businesses with proprietary workflows, customer relationships and physical networks that are hard to replicate.

The market also may be underestimating the duration and size of the opportunity in the AI infrastructure market. AI penetration is still extremely low at like 1% or 2%, and the current aggressive CapEx (capital expenditure) cycle has only gotten us to that point.

If adoption moves materially higher, the world may need far more compute, memory and power, meaning that this story has not fully played out yet.

Finally, the market seems focused on AI replacing white collar office jobs like programmers, but we think it will also affect traditional labour markets as humanoid robot adoption increases over the next decades, enabled by physical AI.

Tye Bousada: Good investors have a well calibrated sense of future regret, and we believe that crowds may be putting too much of their money into a single idea today.

Having a well calibrated sense of future regret means understanding that the world is an uncertain place. History has taught us that no one knows exactly how the future is going to play out and it follows that you're not managing future regret by putting too much of your money into a single idea.

Today, the top ten businesses in the S&P 500 make up 38% of the index. Eight of those top ten ideas are AI themed companies. We believe there's a lot of positive change going on inside the tech space today, and we believe some of that change is giving us the opportunity to buy growth for free.

So, we're investing in some technology businesses, but we don't think putting eight of your top ten weights into a similar idea is wise. Over the years, we've learned that you could be wrong about an idea and when that happens, if you have too much of your money invested in that idea, it leads to very bad outcomes.

Is AI changing where you’re seeking investment opportunities?

Mick Dillon: AI is definitely changing the opportunity set. It's not changing our investment philosophy, which is always we want companies who do something special for their customers. But what's defined as special changes over the time, particularly if say, AI as a tool, changes how you can serve customers with your customer experience.

I'll give you an example. It's natural for industries to change and mature over time. In power supply, data centres are very power hungry, and after 15 years where there was no growth in power supply across grids, we're now seeing huge demand for power coming in and that changes as an example, gas turbines.

There are many different parts of industries that are changed because of AI and yes, that is definitely changing the opportunity set.

Danielle Menichella: It's changing where we're looking, but it's not changing what we're looking for. We still want innovative growth businesses with strong competitive moats, market leadership and attractive business spaces.

AI simply forces us to reassess where those businesses will emerge. Right now, we're taking a selective approach. As the technology matures, we are evolving our exposure across enablers like infrastructure, builders of AI-powered products, and beneficiaries that can use AI to drive revenue growth, efficiency gains or both.

At the same time we're trying to steer clear of obvious losers. That includes businesses where AI may compress pricing, weaken differentiation, reduce barriers to entry, or make the end product easier to replicate. Certain areas of software require real caution.

The key is re-underwriting every company through an AI lens. Does AI improve growth, pricing power, margins, or competitive position? Does it create scarcity or oversupply?

Tye Bousada: AI is obviously a crowd favourite theme these days, and history has shown that crowd favourites lead to high valuations and high valuations lead to lower expected future returns.

Let's look at the S&P 500, because it seems to be where the most excitement seems to be around AI these days. You could go back over the last 35 years and look at the S&P 500’s valuation is measured by its price-to-earnings.

And from those different entry valuations look at what the subsequent ten-year returns have been Based on our figures the data shows that the higher the valuation you pay for something, the lower the real returns over the next decade.

The message is pretty clear: entry price dictates returns. So what does this data tell us about the expected returns when we pay the valuation multiples that the US market is asking us to pay today?

Historically, based on our figures, they've ranged from -4% a year to plus 4% a year for a decade. Not great. I want to be clear we still own US businesses, but we own far less than the indexes do today.

The MSCI World Index, which is a global benchmark has 71% of the index invested in the US, a record high. We have less than half of that in the US.