Our second look at what 2026 might hold for institutional investment teams explores the impact of digital transformation, both in terms of the growth of digital assets and the arrival of artificial intelligence in the world of asset allocation and portfolio management.
It almost seems an understatement that the scope, capabilities and impact the two key drivers of digital transformation are revolutionary. There is no hiding from them but such is the rapid pace of their development that predictions almost seem rash. The counter to that is ducking an assessment of where we are today will just handicap you in shaping where you and your firm want to be tomorrow.
What did our experts, coming from different perspectives, make of this?
Our contributors
Sian Fisher (SF) • Currently a non-executive director of major London Market insurance firms. Chair of the Worshipful Company of Insurers Insurance Non-Executive Directors Forum. Past CEO of the Chartered Insurance Institute. Senior roles at Hiscox and Arthur J Gallagher.
Erik Vynckier (EV) • Former chair of the investment committee at Foresters Friendly Society and investment consultant
David Worsfold (DW) • Contributing Editor, Insurance Investment Exchange
When will the growth of digital assets hit insurer’s investment portfolios?
EV • It’s a bad idea. Digital assets have no track record. I do believe the standard custody practices with duplication in physical records, made available online as needed, is cheaper and certainly more secure.
Digital assets are a cyber incident waiting to happen (unless they are just fake replicas of de facto standard assets). Failure of databases etc. eliminates a golden plate verification of ownership and terms and conditions.
The settlement cycle of assets in custody allows for post-trade checks on the integrity of the transaction and sound verification of all counterparties. It prevents mistakes from crystallising and blocks greater harm. For the time being, I find it would be foolish to do away with these checks, which open the door to widespread operational failures and outright fraud.
SF • Since the basis of investing in these assets is currently “first secure a bigger idiot to sell them onto” unlikely any time soon. If Stablecoin is just existing governments with a more digital appearance, what’s the point.
DW • Extreme volatility is often cited as one of the biggest enemies of successful portfolio management. Digital assets exhibit wild, unpredictable volatility. The nature of their creators and principal investors means this is hard-wired into them. Some say that the entry of institutional investors into the market would bring some stability but who would be the first to risk their portfolio and their reputation in what still looks like a reckless gamble in a wild west saloon?
Their day will come but it is still very hard to see when and how that will arrive.
Where will AI make the biggest impact in your world?
EV • There is no “thinking” or “intelligence” involved in AI. Rather it is brute force automation. The automation is essentially correlative in a very high dimensional space and involves very limited methodological input: that is its attraction and its curse.
Rather it throws computer power and ultra-large calibration datasets at a task in the hope “thinking” will not prove necessary.
As an example, Large Language Models use millions of dimensions and calibrate a neural network with trillions of parameters to hopefully get some success with the correlative approach. The calibration is essentially against the entire internet (of which 70% is of questionable value). It is hit-and-miss but with very many dimensions and even more parameters, it might occasionally get some things right by chance. It takes no responsibility for accuracy and does not attempt to “think”. There is no inherent quality control. There is no learning after the calibration has completed and there isn’t constant updating (as this would invalidate the earlier calibrations for the trillions of parameters on the neural nodes).
I do think that LLMs have broken through in search and are here to stay for now, with warts and all. The output is unreliable and makes no intellectual steps forward (it does no thinking whatsoever) and therefore, everything needs to be proofed and an expert will probably shake his/her head reading the replies of an LLM in his/her field of expertise. But that was already the case with classical “googling” just the same.
SF • One interesting comment on this from a colleague – “We will have massive initial disintermediation of the old ways but this only ever leads to reintermediation in a new way- who will we trust?”
At a prosaic personal level, hopefully non-executive directors will be able to find things across myriad past board packs. And extract the “so what” from the repetition and duplication across multiple functions reports, for clarity of decision making.
DW • Whenever I am asked about the potential of AI, my first thought is always where is the “human in the loop”. This is as important in the world of asset allocation and management as it is in any other area where AI is expected to make an impact.
For most CIOs the challenge is to use AI as an enabler of human decision-making, not a substitute for it. They must establish feedback loops where human expertise and AI insights refine each other. By doing so, these firms create an organisational intelligence greater than the sum of its parts, a mutual beneficial relationship.
Trust in the reliability of AI outputs has held back its deployment – sensibly in most cases. The cautious adoption by some CIOs starts with only using data they already hold and have validated, built into their existing models. Only once they have scrutinised the AI outputs when asked, say, to propose fresh allocation strategies are they slowly introducing external data sources. This is where the power of AI will really start to make an impact.
For now, AI’s primary contribution is its ability to increase analytical capacity, helping investment teams assess portfolios more frequently and efficiently. Given today’s volatile markets, AI’s ability to process structured and unstructured data at scale could significantly improve the quality and frequency of strategic allocation reviews in particular.
Looking ahead, AI's role in institutional investing will expand beyond structured data analysis to include natural language processing (NLP) – this would be a major step change. This evolution will allow AI to process company reports, regulatory filings, central bank statements, and even social media trends, enabling more dynamic investment strategies, including thematic investing.
There are also legitimate concerns about systemic risks from widespread AI adoption. If many firms rely on similar AI models and data sources, there is a risk of herding, where models make the same decisions simultaneously, potentially amplifying market volatility. Again, the role of human expertise and judgement will be crucial in avoiding or mitigating such hazards.