AI Is Eating Asset Management's Middle Layer. Will Your Fees Fall?

The engine room of your fund is being rebuilt, and nobody sent you a notice#

If you own a mutual fund, an ETF or a unit-linked pension, you pay a small annual slice of your money to a company you probably never speak to. That slice buys a fund manager, yes. It also buys an enormous and invisible apparatus: analysts reading filings, dealers placing orders, operations staff matching trades, accountants striking a daily price, compliance officers checking the lot.

That apparatus is the middle layer of asset management. And in 2026 it is being handed, function by function, to software.

Boston Consulting Group's Global Asset Management Report 2026, published on 28 April, estimates that AI could cut industry costs by 25% to 35% over three to five years. BCG's companion analysis, Rebuilding Asset Management for an AI-First World, goes further and puts operational cost reduction in investment operations at roughly 40%, with software agents running fund accounting, reconciliations and corporate actions on their own.

So here is the question that matters to anyone with a SIP, an ISA or a 401(k). If the factory gets cheaper to run, does the product get cheaper to buy?

The evidence so far says: not automatically, and not yet.

What the "middle layer" actually is#

Fund charges are usually quoted as an expense ratio, or total expense ratio (TER): the percentage of your money deducted each year to run the fund. A 0.40% expense ratio means £40 a year on a £10,000 holding, taken quietly from the fund's assets rather than billed to you.

That percentage pays for three broad things. The front office picks investments. The back office keeps records, settles trades and calculates the daily net asset value. Between them sits the middle office, which handles risk, performance measurement, reconciliation and the endless checking that the two ends agree.

A few terms worth knowing before we go on:

Reconciliation is the daily job of confirming that what the fund thinks it owns matches what the custodian bank says it owns. Corporate actions are events such as dividends, splits and mergers that change what a holding is worth and must be processed correctly. An agentic AI system, unlike a chatbot, does not just answer questions; it takes a goal and executes a sequence of steps, calling other software as it goes.

These jobs are repetitive, rule-bound and expensive in people. Which is exactly the profile of work that current AI systems handle well.

The machines are already in the engine room#

This is no longer a pilot project. Some hard numbers from primary sources:

BNY, one of the world's largest custodian banks and the firm that keeps the books for thousands of funds, disclosed in its first-quarter 2026 investor update that it had around 220 enterprise AI solutions in production and about 140 "digital employees", its term for multi-agent systems that work alongside staff. More than 40% of the bank's code in that quarter was written by AI. Expenses rose 5% while revenue rose 13%, and the pre-tax margin reached 37%.

Adoption is broad rather than exotic. Mercer surveyed 131 asset managers between February and March 2026 and found that 55% had integrated AI into at least one investment process. Three-quarters used it for automation and efficiency. Only 6% let it make decisions. SimCorp's InvestOps 2026 study of 200 senior executives at firms managing at least $10bn found 70% using AI to support the front office, against roughly 10% merely exploring it a year earlier.

The staffing consequences are visible in banking, where disclosure is better. American Banker reported in July 2026 that Standard Chartered plans to cut 7,800 back-office roles by 2030 and that JPMorganChase's operations headcount fell 4%, with operations the only job category showing no growth at firms investing heavily in AI.

Here is the part the marketing rarely mentions. In the Mercer survey, 69% of managers said AI had improved operational efficiency, while just 8% claimed it had improved returns. Read that twice. The technology is delivering a cheaper process, not a better portfolio.

Why cheaper to run does not mean cheaper to own#

A fund's fee is not a cost-plus calculation. It is a price, set by what competitors charge and what distributors will accept. Costs determine the manager's margin. They do not determine your bill.

BCG's data makes the point bluntly. Between 2010 and 2025, industry revenues grew 5.1% a year while costs grew 5.4%, and profit margins sat at roughly 30%, essentially unchanged, even as assets under management nearly tripled to $147trn. Scale, in other words, has already delivered vast efficiencies over fifteen years. Margins stayed flat anyway, because savings were competed away in some segments and swallowed by rising costs in others.

BCG's own framing of the AI opportunity is telling. Its AI-first analysis argues that advantage comes from judgment and client relationships rather than from labour savings alone, and that firms can use freed capacity to expand customisation and service while holding margins. That is a strategy for keeping the benefit, not passing it on.

My reading, and this is interpretation rather than fact: where a product is commoditised and price-transparent, such as a broad index ETF, AI savings will reach investors quickly, because a rival will cut first. Where the product is bundled, advised or opaque, such as a multi-asset pension default or a private markets fund, the savings are far more likely to stay with the manager.

There is also an awkward asymmetry. The functions AI automates most easily, operations and reporting, are a modest share of a typical fund's charge. Portfolio management and distribution, the expensive parts, are harder to automate and, in the case of distribution, are precisely where the industry is currently spending more.

Where fees actually fell, and why it was not the robots#

Fund charges have fallen a long way, and the reason is well documented and rather unglamorous.

Source: Investment Company Institute, Trends in the Expenses and Fees of Funds, 2025, published 25 March 2026.

Note what did not happen in 2025. The average equity fund fee did not move. Money market fees went up. The ICI attributes the long decline to three things: a shift to no-load funds, which took 92% of 2025 sales against 46% in 2000; a migration into index funds, which grew from 19% to 52% of long-term fund assets between 2010 and 2025; and investor preference for cheaper options generally.

So the average fell mostly because investors moved their money, not because managers cut prices. Morningstar's 2026 US Fund Fee Study, published on 19 May, puts the asset-weighted average expense ratio at 0.32% for 2025, down from 0.80% in 2006, and estimates investors saved about $6.8bn in fees during the year.

Genuine price cuts do happen, and they come from scale rather than from algorithms. Vanguard announced on 2 February 2026 that it had cut charges on 53 funds across 84 share classes, an estimated $250m of savings in 2026 and $600m since 2025, taking its asset-weighted average to 0.06%. Schwab followed on 11 June, trimming its mid-cap and small-cap ETFs from 0.04% to 0.03% and its emerging markets ETF from 0.07% to 0.06%. Neither firm credited artificial intelligence. Both credited growth and scale.

The rulebook is moving faster than the fee schedule#

Regulators have reached the topic from two directions: policing claims, and forcing costs down directly.

On claims, the US Securities and Exchange Commission has been pursuing what it calls AI washing since it charged two investment advisers in March 2024 over false statements about their use of AI, settling for $400,000 in combined penalties. The then chair, Gary Gensler, said that advisers "should not mislead the public by saying they are using an AI model when they are not". Expect more of this as marketing outruns deployment.

On strategy, the UK Financial Conduct Authority published its review of AI in retail financial services on 6 July 2026, concluding that AI will be a defining force in the sector by 2030 and flagging both efficiency gains and risks around consumer protection and market concentration. Roughly 20% of UK adults, about 11 million people, told its researchers they were likely to use autonomous AI tools to manage their money.

India has taken the most direct route: it simply lowered the ceiling. The Securities and Exchange Board of India's board meeting of 17 December 2025 replaced the old total expense ratio with a base expense ratio covering the fund house's own controllable costs, with statutory levies charged separately on actuals. Caps for index funds and ETFs fell from 1.00% to 0.90%, equity schemes under Rs 500 crore from 2.25% to 2.10%, and permitted cash-market brokerage from 12 basis points to 6. The new framework took effect from 1 April 2026.

That is worth holding onto. In the one large market where charges fell by decree this year, the cause was a regulator, not a robot.

Key takeaways#

1. AI is genuinely automating the middle and back office of fund management. BCG estimates roughly 40% operational cost reduction in investment operations, and BNY already runs about 140 multi-agent "digital employees".
2. Efficiency is not performance. Only 8% of asset managers surveyed by Mercer said AI had improved returns, against 69% reporting better operational efficiency.
3. Fees are prices, not costs. Industry margins have sat near 30% since 2010 despite assets nearly tripling, so past efficiency gains did not reliably reach investors.
4. The fee decline you have read about came mainly from investors switching into cheaper funds, not from managers cutting prices. Average US equity fund charges were flat in 2025 at 0.40%.
5. Where charges genuinely fell this year, the drivers were scale (Vanguard, Schwab) and regulation (SEBI), not artificial intelligence.

Frequently asked questions#

Will my fund's expense ratio drop because the manager uses AI? There is no evidence of that yet. The 2025 data shows average equity fund fees unchanged. Cost savings reach investors when competition or regulation forces the issue.

Is an AI-run fund likely to perform better? The survey evidence is discouraging on that point. In Mercer's 2026 study, only 8% of managers reported improved returns from AI and only 6% used it for decision-making at all.

Where can I see what I am actually paying? The expense ratio or TER appears in the fund's factsheet, key information document and annual report. In India, schemes publish daily TER on the AMC website. Compare the same fund's direct and regular plans, or different share classes, because the underlying portfolio may be identical while the charge is not.

Does a lower fee guarantee a better outcome? No. It guarantees a lower deduction. Cost is one of the few variables you can know in advance, which is why it gets so much attention, but fund choice involves risk, mandate and time horizon too. This article is information, not advice.

If AI cuts jobs in fund operations, is my money less safe? That is an open question rather than a settled one. The FCA's 2026 review flags consumer protection and concentration risk among its concerns, and automation moves staff from doing the work to supervising it. Nobody has good data yet on error rates at scale.

How do I spot a firm exaggerating its AI credentials? Look for specifics: which function, deployed since when, with what measurable result. The SEC's 2024 enforcement action turned on advisers claiming capabilities they did not have.

Does any of this apply to small investors in India? Directly. SEBI's new base expense ratio framework, effective 1 April 2026, lowered the caps on what a scheme can charge and pulled statutory levies out of the headline number.

Glossary#

Expense ratio / total expense ratio (TER) The annual percentage of your investment deducted to cover the fund's running costs.

Base expense ratio (BER) SEBI's new measure of the fund house's own controllable costs, separate from statutory levies such as GST and stamp duty.

Middle office The layer between investing and record-keeping: risk monitoring, performance measurement and reconciliation.

Reconciliation Daily checking that the fund's own records of holdings and cash match the custodian's.

Corporate actions Events such as dividends, bonus issues, splits and mergers that must be processed accurately in a fund's books.

Agentic AI Software that pursues a goal through a multi-step sequence, calling other systems as needed, rather than simply responding to a prompt.

Asset-weighted average fee The average charge weighted by how much money sits in each fund, which reflects what investors actually pay rather than what is advertised.

AI washing Marketing a product as AI-driven when the underlying capability is absent or overstated. The SEC treats it as a disclosure violation.

References#

1. Boston Consulting Group, Competition in Asset Management Shifts as Growth Becomes Harder and More Concentrated, 28 April 2026.
2. Boston Consulting Group, Global Asset Management Report 2026: Rebuilding Asset Management for an AI-First World, 2026.
3. Boston Consulting Group, An Imperative for Growth and the New Economics of Asset Management, 2026.
4. BNY, Quarterly Update Presentation, 1Q 2026, 16 April 2026.
5. Mercer, Moving Beyond the AI Pitch: Asset Managers' Use of AI, 2026.
6. SimCorp, More Than Two-Thirds of Investment Managers Prominently Using AI to Support Front Office, 19 January 2026.
7. Investment Company Institute, Trends in the Expenses and Fees of Funds, 2025, 25 March 2026.
8. Investment Company Institute, Mutual Fund and ETF Fees Remained Near Historic Lows in 2025, 25 March 2026.
9. Morningstar, 2026 Annual US Fund Fee Study, 19 May 2026.
10. Vanguard, Vanguard to Deliver More Than Half a Billion in Expected Savings to Investors Since 2025, 2 February 2026.
11. Charles Schwab, Schwab Asset Management Reduces Fees on Four Equity Index ETFs, 11 June 2026.
12. US Securities and Exchange Commission, SEC Charges Two Investment Advisers with Making False and Misleading Statements About Their Use of Artificial Intelligence, 18 March 2024.
13. Financial Conduct Authority, FCA Publishes Landmark Review into Impact of AI on Retail Financial Services, 6 July 2026.
14. Securities and Exchange Board of India, SEBI Board Meeting, PR No. 84/2025, 17 December 2025.
15. Cafemutual, SEBI's New Expense Ratio Framework: What Changes After April 1, 2026, 2026.
16. American Banker, AI Is Starting to Replace Humans in Operations, Analyst Jobs, 8 July 2026.