AI is already part of many asset managers’ investment workflows, but the headline figures are easy to misread. Mercer’s 2024 global manager survey found 54% were using AI for investment strategy or asset-class research and another 37% planned to use it. That produced the often-repeated 91% figure, but it was a combined current-use-and-intention measure—not a claim that 91% were already using AI. Mercer’s later survey of 131 asset managers, conducted in February 2026 and released May 21, 2026, found 55% had integrated AI into at least one investment process, 27% were running pilots or proofs of concept, and 18% had not integrated it. A separate 91% in that survey meant firms planned to increase AI use during the following 12 months.
The frequently quoted $14.82 billion market estimate should not be treated as established fact. The figure appears in a secondary online document, but an authoritative original publisher, market definition and methodology were not identified. The reliable story is about how firms are deploying AI, where humans still control decisions, and what governance weaknesses can undermine the benefits.
As an Amazon Associate I earn from qualifying purchases.
What the 91% figure actually measures
Mercer’s 2024 global manager survey examined AI use in investment strategy and asset-class research among managers represented in its Global Investment Manager Database. It reported three broad groups: 54% already using AI, 37% planning to use it, and the remainder not reporting current use or a stated plan in the measure described. Adding the first two groups gives 91%.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11That arithmetic matters. “Using or planning to use” is not the same as “using.” A manager with a budget request, a scheduled pilot or a multi-year roadmap is at a different implementation stage from one running a model in a live research process. The 37% also covered plans over different horizons, so the combined figure does not identify when each planned deployment would occur.
#1 Best Overall
The survey scope is another qualification. These were responses from investment managers in Mercer’s database, not an audited census of every asset manager in every country. It is best read as an indication of direction among surveyed institutional managers rather than a universal adoption rate.
Why headlines turn intention into adoption
Adoption statistics often combine several stages because surveys ask whether an organization is using a technology now or expects to use it. That is useful for measuring momentum, but it conceals the difference between experimentation, production use and an announced intention. Any article, presentation or vendor claim using the 91% number should state the full wording and the 2024 survey year.
What Mercer’s 2026 survey adds
Mercer’s February 2026 survey of 131 asset managers, released May 21, 2026, separates implementation status from future plans:
| Measure | Result | What it means |
|---|---|---|
| Integrated into at least one investment process | 55% | Respondents reported live integration somewhere in an investment workflow. |
| Pilot or proof-of-concept stage | 27% | AI was being tested, but was not necessarily operating as a production process. |
| No integration yet | 18% | Respondents had not integrated AI into an investment process at the time of the survey. |
| Plan to increase AI use in the next 12 months | 91% | An intention measure, separate from current integration status. |
The 2026 results provide a more operational picture than the earlier 91% statistic. They show that a majority of respondents had at least one integrated use, while more than two in five were still at pilot stage or had no integration. The 91% intention-to-increase figure indicates strong forward momentum, not that 91% had production systems.
Rank #2
- Ideal for Gifting
- Ideal for a bookworm
- Compact for travelling
These are respondent-reported findings. Mercer did not present them as an independent audit of each firm’s models, data pipelines or trading permissions. A firm may classify a narrow internal tool as “integrated” even when AI has no authority over portfolio construction or trades.
Is AI making investment decisions or assisting people?
Most reported use is assistive rather than autonomous. In Mercer’s 2024 findings, more than half of teams with AI integration said AI analysis informs final decisions rather than determining them. One fifth said AI proposes investment decisions that the team can override. Those categories describe different control arrangements: an analyst may use a model’s summary as one input, while a committee may receive a model-generated proposal but retain the ability to reject it.
Mercer’s 2026 release reported that only 5% of firms currently grant AI autonomous or semi-autonomous authority for investment recommendations or trades. That is a much narrower claim than using AI for research, text analysis or workflow automation. It also explains why high adoption percentages do not imply that machines are independently managing most portfolios.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →“AI is delivering measurable efficiency and insight for asset managers today, but the technology is largely a partner rather than a decision‑maker,” said Beverley Sharp, Mercer’s Global Manager Research Leader.
Human oversight does not eliminate risk. An investment committee can over-trust a confident-looking model, fail to notice stale data or approve a recommendation it cannot explain. The meaningful question is therefore not simply whether a firm uses AI, but where the system sits in the decision chain, what data it can access, and who is accountable for the final action.
Where asset managers are using AI
Mercer describes applications across research, analysis, idea generation and decision support. Typical workflows include:
- Research retrieval: searching large collections of filings, transcripts, reports and internal notes to locate relevant passages faster.
- Information synthesis: summarizing documents, comparing disclosures and highlighting changes for an analyst to verify.
- Idea generation: proposing themes, securities or questions for further investigation rather than issuing an unreviewed order.
- Decision support: organizing signals, scenarios and supporting evidence for an investment committee.
- Operational work: automating repetitive classification, data preparation, monitoring and reporting tasks that previously consumed analyst time.
These uses vary in consequence. A drafting assistant that produces a first-pass summary has a different error profile from a model that feeds a portfolio-constraint engine. Controls should become stricter as an output moves closer to a recommendation, allocation or trade.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What benefits have firms reported—and what the surveys do not prove
In the 2026 release, respondents most clearly cited operational efficiency and faster or higher-quality insights. Wider datasets, quicker analysis and more ideas can reduce manual work and allow teams to examine more information. Those are useful business outcomes, but they are not the same as investment performance.
Rank #4
Neither Mercer survey establishes that AI adoption improves returns, lowers volatility or consistently produces alpha. A faster research cycle may help a team, yet a model can still misinterpret a filing, amplify a crowded trade or rely on data that is incomplete. Efficiency is an input to an investment process, not proof of a superior result.
| Claim | Supported interpretation | Unsupported leap |
|---|---|---|
| More firms are integrating AI | AI is becoming a normal component of surveyed workflows. | Every firm has a mature, reliable system. |
| Faster insights | Respondents report speed or quality improvements in analysis. | Those improvements guarantee higher returns. |
| AI proposes decisions | People receive model-generated recommendations they can review or override. | AI controls portfolios without human responsibility. |
Why implementation is difficult
Mercer identifies data quality and availability, integration and compatibility, and ethical and legal considerations as major barriers. The problems are connected: a model cannot produce dependable analysis from inconsistent identifiers, missing histories or licensing-restricted data, and a technically sound model can still create unacceptable privacy or conduct risks.
Data quality and availability
Investment data arrives in different formats, with revisions, survivorship issues, time-zone mismatches and gaps. Firms need documented ownership, validation checks and a way to trace an output back to the underlying source. Generative systems also require controls against fabricated citations or plausible-sounding but unsupported statements.
Integration and compatibility
A pilot in a sandbox is easier than connecting a model to research libraries, portfolio systems, order-management tools and identity controls. Interfaces, permissions, latency and version changes can all break a workflow. A production process needs monitoring, rollback procedures and a clear owner when an upstream system changes.
Best Value
- It can be a gift option
- Comes with secure packaging
- Helpful in various ways
Ethical, legal and conduct questions
Confidential client information, material nonpublic information and licensed research cannot be sent to an uncontrolled service. Firms must define retention, access, auditability and disclosure rules. Mercer’s recommendations include governance, data privacy and security, awareness of model limitations, and realistic planning for costs and skills. These are Mercer’s implementation recommendations, not a universal legal checklist; applicable requirements depend on jurisdiction, activity and the firm’s regulators.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.A practical framework for evaluating an AI investment workflow
An adoption percentage says little about quality. Investors, boards and clients assessing a manager can ask the following questions.
- Identify the exact task. Is AI summarizing documents, ranking ideas, forecasting a variable, setting a constraint or sending an order? Describe the workflow rather than labeling the whole firm “AI-powered.”
- Classify the operating stage. Separate production integration from a pilot, proof of concept or future intention. Ask when the system went live and how often it is used.
- Map decision authority. Determine whether a person must approve each recommendation, whether overrides are recorded, and whether any system can place or amend a trade.
- Test the data chain. Establish where inputs originate, how freshness and completeness are checked, and how the firm handles conflicting or missing information.
- Evaluate the model. Review validation methods, known failure modes, performance across market regimes and procedures for detecting drift after deployment.
- Inspect governance. Look for access controls, privacy protections, vendor and model inventories, audit logs, incident reporting and documented accountability.
- Measure the right outcome. Track time saved, error rates, review quality and decision latency separately from portfolio returns and risk. Do not treat an operational metric as a performance guarantee.
How to read the $14.82 billion market claim
A secondary online document repeats a $14.82 billion estimate for an applied-AI-in-finance market in 2025. The available material does not identify an authoritative original publisher, the geographic coverage, the companies included, the revenue categories counted or the calculation method. “AI in financial services,” “AI trading,” “applied AI in finance” and enterprise software for financial firms can describe materially different markets.
Without a primary source and methodology, $14.82 billion is an unverified estimate, not a dependable market size. It should not be used to calculate a firm’s market share, justify an investment thesis or support a forecast. A credible market figure would specify the base year, currency, region, inclusion rules, data sources, forecast period and whether the value represents vendor revenue, spending or projected opportunity.
What these numbers mean for individual investors
Most individual investors will encounter AI indirectly through a fund manager, brokerage feature or financial-planning service. The Mercer figures show institutional experimentation and deployment, not that an AI label makes a product safer or more profitable.
- Ask whether a tool provides information, a recommendation or an automated action.
- Check whether a qualified person reviews outputs and whether you can see the assumptions behind a recommendation.
- Be cautious with confident forecasts that do not identify data sources, time horizons or uncertainty.
- Never share account credentials, tax documents or other sensitive information with an AI service unless its security, retention and authorization practices are clear.
- Judge a service by transparent costs, diversification, suitability and risk controls—not by an adoption statistic alone.
The clearest conclusion from the adoption data
AI has moved beyond isolated experiments for many surveyed asset managers, but adoption is layered. Mercer’s 2024 91% combined current-use-and-plan figure measures momentum; Mercer’s 2026 results show 55% integrated, 27% piloting and 18% not yet integrated, with 91% planning to increase use. At the same time, only 5% reported giving AI autonomous or semi-autonomous authority for investment recommendations or trades.
The evidence therefore supports a measured conclusion: AI is primarily an increasingly capable research, analysis and operations partner, while human teams generally retain responsibility for investment judgment. Reported efficiency and insight gains are promising, but they do not establish better returns. The quality of data, integration, governance and oversight will determine whether adoption creates durable value.
Recommended Free Tools
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
