T3 Research  ·  AI Index Series  ·  2025 Edition

AI IN
ASSET MANAGEMENT
INDEX 2025

A comprehensive analysis of AI adoption, investment trends, use cases, performance impacts, and regulatory considerations across the global asset management industry. Data through Q3 2025.

ChatGPT dominates the advisor channel at 35.7% - ahead of Microsoft Copilot (12.1%) and Google Gemini (6.9%). But in the enterprise layer, Claude is climbing fast: 70% of Fortune 100 companies now use it, with a 29% enterprise market share in 2025. Integration with Teams, Slack, and Zoom drove a 72% year-over-year increase in business usage. The two stories together reveal a split market: consumer-facing advisory standardises on ChatGPT workflows, while institutional back-ends increasingly embed Claude for compliance-sensitive, document-heavy enterprise tasks.

GenAI is big enough to matter - but 70% of AI workloads in asset management are still traditional machine learning: factor models, risk engines, and quantitative analytics. Agentic AI is emerging at 7%, with autonomous systems managing complex tasks end-to-end. Meanwhile, 87% of firms use closed-source LLMs, driven by vendor support guarantees and compliance requirements rather than model quality alone. The implication: firms need governance frameworks that cover both legacy ML and the fast-growing GenAI / agentic layer.

Portfolio management support is in the top tier of deployed use cases - not pilots. Customer analysis and personalisation lead, confirming the signals-and-research layer is operational.

Implementation momentum often starts in engineering and operations, not the investment desk. Code generation and knowledge management deliver quick wins before more complex front-office deployments.

The gap between intent and implementation is stark: three-quarters of organisations acknowledge AI testing is strategically important, yet only around one in six have put formal testing practices in place. For asset managers operating under fiduciary obligations, this disconnect represents both a compliance risk and a competitive opening. Firms that close this gap - with bias testing, red-teaming, and model validation - will build trust with regulators and clients alike.

Robo-advisors now manage $1.97 trillion globally, led by Vanguard ($311.9B), Empower ($200B), and Schwab Intelligent Portfolios ($80.9B). But the success is concentrated in passive, low-cost distribution: ETF allocation, automatic rebalancing, and tax-loss harvesting for retail clients. Hybrid models that combine automation with human advisor access now hold 45% of market share - confirming that robo-advisory works best as a scalable distribution layer for standardised products, not as a replacement for active portfolio management or institutional decision-making. Top-performing platforms like SoFi (14.08% 1-year return) show competitive results, but primarily in passive index strategies.

The constraint on AI value in asset management is people, not technology. Two-thirds of firms report only small or moderate AI returns, and the top three barriers are all human factors: cultural resistance (55%), poor data quality (51%), and talent gaps (45%). The firms pulling ahead - those reporting 7%+ ROI - invest systematically in apprenticeships, AI-literate hiring, and continuous training programmes. Models are commoditising; the scarce input is the workforce that can deploy them responsibly.

Source: Stanford HAI AI Index 2025

US private AI investment is nearly 12× China's and 24× the UK's. Global GenAI investment reached $33.9B, up 18.7% YoY. This explains why most global asset managers adopt US-anchored model stacks first, then localise for sovereignty.

Cross-border data, client confidentiality, and model residency requirements are now shaping platform architecture and vendor selection - not only security teams.

Asset managers cannot wait for final rules. The consultations are setting the practical compliance burden now - from QMS standards to incident-reporting timelines. Firms that engage early with the evolving definitions will shape their own obligations; firms that wait will inherit them.

Sources: IntuitionLabs (Feb 2026), TLDL (Feb 2026), official provider documentation. Prices verified Feb 2026.

The pricing spread across models is now 1,000× - from $0.05/M tokens (Llama 4) to $168/M (GPT-5.2 Pro output). For asset managers, this means three distinct cost tiers are emerging: premium reasoning models for complex legal and investment analysis ($5-$168/M), mid-tier workhorses for research and compliance ($1-$15/M), and budget models for high-volume operational tasks ($0.05-$2/M). The strategic question is no longer “can we afford AI?” but “are we routing tasks to the right model tier?”

The Data Is Clear. The Leaders Are Pulling Away.

T3 helps asset and wealth managers move from signal to strategy - with regulation-first AI governance, implementation, and training designed for the financial services operating environment.

Or email us at contact@t-3.ai

Sources & Methodology

ThoughtLab / Grant Thornton "The AI-Powered Investment Firm" (Q3 2025, n=500, $74.2T AUM) • Mercer Manager Survey (2025) • EY Wealth & Asset Management GenAI Survey (2025, n=100) • Stanford HAI AI Index Report 2025 • Deloitte M&A GenAI Survey (H1 2025) • KPMG Global PE Outlook (Q3 2025) • Bain Private Equity Report (2025) • AFME EU GPAI Code Submission (2025) • Anthropic Enterprise Data (2025) • Precedence Research (2025) • Wipro Limited • Statista Robo-Advisor Market (2025).

Where industry data does not cleanly exist, the closest defensible proxy is used and called out explicitly. This page is for informational purposes and does not constitute investment advice.