How the Biggest Investors Are Finding Opportunity in AI
“AI’s capabilities may be advancing exponentially but adoption is advancing institutionally, which means years, not quarters. That gap is where much of today’s value is being created — and lost. ”
The recent release by Anthropic of a model that can spot cyber vulnerabilities faster than most humans highlights a growing reality: AI innovation is outpacing the ability of institutions to keep up.
Asset owners and managers daily confront breathless headlines (and ad copy) telling them agents will soon run portfolios and outperform human managers.
The reality is more measured. For all its promise, agentic AI — technology that can act independently — is not yet ready to run your portfolio. Yet the investment industry is still being rewired by AI, slowly and unevenly, but inexorably. AI’s capabilities may be advancing exponentially but adoption is advancing institutionally, which means years, not quarters. That gap is where much of today’s value is being created — and lost.
From our work with many of the world’s largest asset owners, managers and financial institutions, three ideas stand out for those looking to integrate AI in ways that serve their investment goals and risk-management needs.
For investment organizations, most of the immediate value comes from operational gains. Tasks that once took an analyst a week, like extracting data from documents, can now take minutes. Portfolio managers can start the day with a narrative summary of diagnostics instead of waiting for reports. Researchers can synthesize thousands of filings that no individual could fully absorb. The hype is that the agent can replace the analyst or portfolio manager. The reality is an analyst who quietly does the work of five and the manager who uses AI to materially improve decision quality.
For investors, these gains may not look like traditional alpha, but they are measurable, and they compound. Every clean input improves every downstream model. Firms that build AI fluency into their processes now, through structured data, consistent taxonomies and auditable workflows, will be significantly better positioned by the time their peers catch up because they’ve been practicing. It’s why MSCI invests in developing these capabilities alongside our clients, so that we can help them narrow the gap and derive greater value they derive from our data and methodologies.
The most durable edge today sits in the data. We all know that AI models hallucinate. A general purpose model, for example, might produce incorrect estimates for a meaningful share of a company’s climate data. It’s a real and practical constraint.
Yet the harder problems in data aren’t about cleaning spreadsheets. They are semantic and contextual. Investors need to determine what a company is actually doing from disclosures that weren’t designed for machine interpretation, let alone to answer specific investment questions.
We see the challenge everywhere we work. In private markets, general partners deliver reports in PDF format with taxonomies that are rarely the same. In front offices, holdings data must be matched, validated and corrected before it can be analyzed. In climate, investors try to reconcile scientific risk models with corporate disclosures and regulatory frameworks that seldom align. As a data provider to large institutions, we make certain the data our clients’ AI models use can be traced to its source, with its accuracy and provenance validated by a human.
“Layer the most capable frontier model atop inconsistent holdings data, misclassified reports or unresolved entity records, and you will get a confident, fluent and wrong answer faster than ever.”
That pairing is the point. None of this is a mechanical exercise. It's research. Layer the most capable frontier model atop inconsistent holdings data, misclassified reports or unresolved entity records, and you will get a confident, fluent and wrong answer faster than ever.
Earlier this year, the MSCI Institute developed a large language model to identify listed companies building the adaptation economy. The challenge is definitional. A company producing weather-resistant roofing, drought-tolerant seeds or grid-hardening equipment rarely describes itself as a resilience company. The language companies use is industrial, not thematic.
We built a methodology that processes roughly 900,000 tokens per company, including annual and sustainability reports, product descriptions and earnings transcripts, and the picture changed entirely. Across about 2,500 companies representing roughly USD 96 trillion of market value, 89% showed evidence of at least one resilience-related activity, and roughly half generated revenue from products that enable resilience. The information wasn’t hiding, it was obscured by vocabulary.
The approach points to broader changes in investing. Traditionally, investors were limited by the time available to define a theme and construct a universe. That imposed a ceiling. Methodologies such as those we’ve been refining raise that ceiling and make it dynamic. A thematic universe can not only be explored, but also recomputed as new information emerges rather than updated on a fixed annual cycle.
“Research is not being replaced, but the infrastructure that underpins it is suddenly usable at a speed, depth and scale that allows investors to answer the questions they want to ask.”
That's the real shift underway. Research is not being replaced, but the infrastructure that underpins it is suddenly usable at a speed, depth and scale that allows investors to answer the questions they want to ask. Signals that once required a full research cycle can now emerge in weeks. Investors can tune themes to their own house view — narrower, broader, regionally weighted, differently benchmarked — and see the universe reshape around their choice, with the methodology documented and auditable at each step.
The shift opens up possibilities for investors across the spectrum. The biggest asset owners and managers can create bespoke investment universes and products at a fraction of the time and cost than was previously possible. At the same time, AI is making such exposures accessible to whole new swaths of mid-sized asset and wealth managers who have not had access to the universe of institutional-grade data and analytics until now. As the cost of analytical customization collapses, consumption of such customization rises by an order of magnitude, recreating Jevons’ paradox for the age of AI, in our view, in the form of an ever-increasing hunger for more bespoke investment solutions.
None of this works without the thematic frameworks, the factor and analytical models, and the entity resolution built over decades — the institutional substrate that MSCI provides. AI acts as a force multiplier on that foundation, not a shortcut. For investors who understand that, the gap is no longer a constraint. It's an edge.
Subscribe todayto have insights delivered to your inbox.
The Hidden Adaptation Economy: A New View of Corporate Resilience and Opportunity
How investors can use an AI-assisted methodology to identify which firms are building resilience to physical climate risks and generating adaptation-related revenues.
Why Wealth Managers Should Think Differently About AI
Advisers may benefit most from AI when they apply it to core competences such as proposal generation and personalized offerings, measuring progress against their own goals.
AI Portfolio Insights
Scale your risk-management process and act faster with GenAI and modern data warehousing.
The content of this page is for informational purposes only and is intended for institutional professionals with the analytical resources and tools necessary to interpret any performance information. Nothing herein is intended to recommend any product, tool or service. For all references to laws, rules or regulations, please note that the information is provided “as is” and does not constitute legal advice or any binding interpretation. Any approach to comply with regulatory or policy initiatives should be discussed with your own legal counsel and/or the relevant competent authority, as needed.