How AI is reshaping the funds industry: what is real, what is hype, and what comes next

Back to Insights

Fund managers are data businesses that happen to manage capital. That statement would have sounded like provocation five years ago. Today it is closer to a job description. The firms that have recognised it are pulling ahead in ways that will be difficult for the rest to reverse. The firms that have not are running increasingly complex portfolios on infrastructure designed for a simpler era.

I spent 18 years inside HSBC Alternative Investments, building the operational infrastructure that scaled the business from $200m to $31.6bn across hedge funds, private equity, real estate, and diversified lending. Zero FCA regulatory breaches across multiple compliance cycles. That experience taught me something specific about the relationship between data infrastructure, operational discipline, and commercial outcomes. The same principles apply now, at greater scale and with more powerful tools. The question for every fund manager is whether they are building the capability fast enough to use what they already have.

What is real

BlackRock runs Aladdin, a risk and portfolio management platform with approximately $25 trillion in assets sitting on it. That is the single largest piece of investment infrastructure in the world. It started as a risk management workstation in a one-room office and evolved into the operating system of the world's largest asset manager. That trajectory tells you everything about what happens when data infrastructure is treated as a strategic investment rather than a cost centre.

Man Group has been running machine learning models in live client portfolios since 2014 through Man AHL and a research partnership with the Oxford-Man Institute at the University of Oxford. Over 70% of the firm's people now use generative AI in their daily work. Their core analytics platform, Condor, was rebuilt so that multi-asset research calculations that previously took 12 hours now complete in 30 minutes. That is not a marginal efficiency gain. It is a structural change in how quickly the firm can act on its own analysis.

Two Sigma manages over $70 billion and describes AI as the operating system for how its quantitative research and investing work. Their 2026 outlook stated it plainly: the future is not AI replacing humans, it is humans who use AI well replacing humans who do not. That is the co-founder of a $70 billion fund saying it publicly, and the evidence across the industry supports the claim.

Bridgewater Associates has built what it calls an artificial investment associate that operates across portfolio management, trading, and risk management. Schroders reported in March 2026 that AI tools are now part of the daily workflow for hundreds of analysts and portfolio managers across equity and credit desks. Morgan Stanley's AI Assistant achieved 98% adoption across 16,000 financial advisors, and the firm attributed $64 billion in net new assets in a single quarter partly to the efficiency gains.

These are not experiments. They are how these firms run.

What is hype

The idea that AI replaces the portfolio manager or the fund administrator is not supported by what is actually happening at the firms using it most aggressively. Schroders describes 2026 as the year AI moves from a productivity tool to investment insight. That is enhancement, not replacement. The human with the judgement and the regulatory accountability is still in the room. The work changes. The accountability does not.

The claim that every fund manager needs generative AI right now is equally misleading. Most fund managers do not have the data infrastructure to make generative AI useful. Without clean, governed data pipelines, well-architected storage, and reliable reconciliation, you are feeding inconsistent data into an expensive system and getting plausible-sounding outputs that cannot be trusted in production. The engineering foundation comes first. The AI comes second. Firms that reverse that sequence end up with impressive demonstrations and unreliable operations.

And the assumption that AI adoption is cheap and clean deserves more scrutiny than it typically receives. Google used enough cooling water in 2023 to supply London with all its water for nine days. Microsoft's datacentres consume as much electricity as Denmark. Ireland's datacentres already account for 22% of the country's metered electricity consumption, more than every home combined, and grid constraints have forced tighter controls on new connections in Dublin. The infrastructure cost of running AI at scale is real, growing, and largely unaccounted for in most firms' business cases. Fund managers making long-term commitments to AI-driven operating models need to understand what they are buying into, not just what it produces.

There is a harder question underneath this that the funds industry has not yet confronted. Many of the same firms scaling their AI infrastructure are marketing ESG-compliant portfolios and publishing sustainability commitments to investors. The operational energy and water footprint of the AI systems they are building sits in direct tension with those commitments. A fund manager running predictive analytics on a cloud platform powered by a datacentre consuming more electricity than a small country, cooled by water drawn from drought-stressed regions, while simultaneously reporting to LPs on the environmental credentials of the portfolio, has a coherence problem that will not stay invisible for long. Regulators, investors, and the firms' own sustainability teams will eventually ask the question. The ones that have thought about it before it is asked will be in a stronger position than those caught without an answer.

Where AI is changing the outcome

In the front office, the shift is from manually assembling information to validating AI-generated analysis. Schroders has hundreds of analysts using AI tools daily across equity and credit. Morgan Stanley's AskResearchGPT synthesises 350,000 documents for institutional teams in seconds. Man Group's Condor platform cut multi-asset research from 12 hours to 30 minutes. In alternatives, LP behaviour modelling is predicting redemption risk before it materialises. Portfolio company early warning analytics are drawing on financial, operational, and market signals simultaneously. Deal sourcing and market mapping are using alternative data at a scale that was operationally impossible five years ago.

In the middle office, AI is compressing the burden that sits at the heart of fund administration. Transaction monitoring using machine learning finds patterns that rules-based systems miss: structuring, layering, and velocity anomalies that only become visible across large datasets. KYC and AML pattern recognition is moving from periodic review to continuous monitoring. The specific cost that AI addresses most directly is the 80% of effort that fund administrators currently spend on data preparation before any analysis begins. At HSBC, we built a data reconciliation model that enabled four staff to manage over 400 Swiss portfolios worth $9 billion, reducing reconciliation breaks from 17,000 to zero. That was not branded as AI. It was data engineering solving the same problem that AI now solves at greater speed and scale.

At board level, the technology is not the point. The board needs to be able to answer three questions when a regulator or investor asks. First, can you explain how AI is governed across the firm and who is accountable when something goes wrong? Second, do you have the people to run an AI-native operating model or are you dependent on a handful of individuals? Third, what is the commercial return on the AI investment and how do you measure it? Man Group openly publishes its AI adoption rate. Morgan Stanley attributed $64 billion in net new assets in a single quarter partly to AI-driven efficiency. If your board cannot point to something equivalent, that is not a technology gap. It is a governance gap.

What happens next

Tokenisation adds further pressure. BlackRock's BUIDL fund sits at $2.5 billion in assets under management. Settlement is compressing from days to seconds. When the pace of execution accelerates, the skills required for oversight must accelerate with it. The talent that can govern real-time, AI-driven, tokenised fund infrastructure does not exist in sufficient numbers today. That is not a problem for 2030. It is a problem for this year's hiring plan.

The firms that will lead the funds industry over the next decade are not the ones with the most data or the largest technology budget. They are the ones that have built the operating model, the governance, and the skills to turn data into decisions at a speed and quality that their competitors cannot match. Man Group, Two Sigma, Schroders, Morgan Stanley, and BlackRock have already answered whether to adopt AI. The question for everyone else is whether your operating model, your people, and your governance are ready for what comes next. That is this quarter's problem, not next year's planning cycle.