Will AI Agents Strengthen Financial-Data Firms’ Pricing Power?

The signal

Agent distribution and recurring-data growth are coexisting. A causal AI pricing premium remains unmeasured.

The interface can move while the licence stays

A financial professional may begin a research task in an AI assistant instead of a dedicated terminal. That changes the visible point of interaction, but it does not tell us who owns the underlying data, who is entitled to receive it or how the supplier gets paid. The data relationship and the user interface are separate economic layers.

LSEG’s December 2025 announcement described a connector that would make information accessible to customers using their LSEG licensed credentials. S&P Global’s February 2026 ChatGPT announcement similarly described access to licensed data through its connector. MSCI’s August 2026 Gemini Enterprise announcement referred to existing entitlements and controlled access. ^[1][2][3]

Those terms support a precise conclusion: the announcements describe new distribution routes for licensed information. They do not describe a universal replacement of data subscriptions by a general AI subscription. Nor do they disclose the price of every customer’s arrangement. A distribution partnership can protect existing revenue, generate new fees or do both; the access design alone cannot identify the split.

LSEG: strong adoption within a selected customer group

LSEG’s October trading update reported that 95% of its top 200 customers used its AI products and more than two-thirds of its top 50 had an MCP connection. Subscription businesses grew 6.6% in the third quarter on the stated constant-currency basis; Data & Analytics grew 5.2%. Management expected a growing revenue contribution from AI over time. ^[4]

This is stronger evidence than a connector launch alone because it includes customer uptake. It also has a defined population. The largest customers are not interchangeable with every customer, and an MCP connection does not establish how many employees use it, how often they query data or how much incremental revenue it produces.

The distinction prevents two opposite mistakes. We should not dismiss the adoption figures as merely a product announcement. We should also not apply the top-customer adoption rate to the entire customer base or multiply it by a hypothetical fee. The disclosed growth and uptake can coexist without telling us the causal contribution of one to the other.

For investors, the useful next step is a consistent cohort: customers with established agent access, their renewal spending, and comparable customers without it. That would show whether easier access increases the value of the information relationship rather than simply moving existing activity to a new screen.

FactSet: recurring value is growing, but the profit test remains

FactSet’s production MCP announcement followed a beta involving 45 firms and more than 800 institutional users. The company described direct access to its datasets for enterprise and agentic applications. That provides a dated distribution event with a stated early user base, rather than evidence that every client had adopted the product. ^[5]

For fiscal 2026, FactSet reported organic annual subscription value of $2.5682 billion, up 7.0%, and annual ASV retention above 95%. It said AI-solutions ASV added during the year more than doubled. Yet its fourth-quarter adjusted operating margin was 33.0%, compared with 33.8% a year earlier. ^[6]

The combination matters. Growing recurring value and retention support commercial resilience. They do not establish that AI has already raised margins or that the increase in AI-related additions is large relative to the entire business. Without the absolute base and cost allocation, neither the contribution to total growth nor the return on AI investment can be isolated.

The margin comparison should not be used to claim that AI caused lower profitability. Costs, investment timing and business mix may all matter. It instead blocks a stronger claim: a positive AI sales indicator does not, by itself, establish improved shareholder economics.

MSCI: market-linked fees can dominate the headline

MSCI reported third-quarter revenue growth of 12.4%, with recurring subscription revenue up 7.7% and asset-based fees up 24.9%. Of the $98.6 million revenue increase, $49.1 million came from higher asset-based fees. Organic recurring-subscription run-rate growth was 8.7%, while reported retention was 95.0%. ^[7]

Our simple bridge shows that asset-based fees accounted for about 49.8% of the dollar revenue increase. That is a revenue-composition calculation, not a claim that the remaining half was caused by AI. It demonstrates why an investor cannot use the total company growth rate as a clean measure of agent-driven subscription demand.

The previously announced Gemini connection is relevant to distribution. The financial results are relevant to the business model. Joining the two is useful, but the timing is not a causal experiment: the connection launched late in the quarter, and the release does not identify a matched group of customers whose spending changed because of it.

This example also explains why recurring-data companies should not be ranked with one blended growth measure. Subscription fees, transaction activity and assets linked to an index respond to different drivers. A strong financial result can be real without answering the narrower pricing question.

S&P Global: distribution and ownership are separate choices

S&P Global reported Market Intelligence revenue of $1.263 billion in the third quarter, against $1.183 billion a year earlier, a rounded 7% increase. It also announced a strategic review of Capital IQ and associated data businesses, with the perimeter and potential outcome still under review. ^[8]

The company’s earlier ChatGPT announcement establishes that its data can be distributed through another interface. The subsequent review establishes that management is evaluating ownership and investment choices. Neither statement, by itself, establishes that AI is destroying the economics of financial data.

A separation could reflect a judgment about focus, investment needs or organizational structure. It could also change which financial series investors need to follow. The analytical priority is to preserve comparable business boundaries if a transaction happens, rather than read a future ownership decision backward into today’s revenue as proof of disruption.

The countercase remains relevant: customers may become less attached to a particular interface when they can query multiple sources through one assistant. A data provider can retain essential content while losing some control over workflow, bundling and the renewal conversation. That possibility needs customer-spending evidence, not an inference from a strategic-review headline.

Who captures the savings from easier access?

Three outcomes are plausible. First, the provider captures more value because its information becomes easier to use across more workflows. Second, the customer captures most of the benefit through fewer manual steps and broader use under an existing licence. Third, the interface provider gains bargaining power by becoming the default place where users work.

These outcomes can coexist. A data owner might retain enterprise revenue while individual seat counts decline. A customer might pay for a broader dataset while spending less on a dedicated interface. A higher usage count could accompany either higher or lower effective revenue per task. This is why usage, licences and price need to remain separate columns.

The strongest bullish case is that differentiated data becomes more useful as the cost of asking questions falls. The strongest skeptical case is that easier comparison and centralized interfaces make customers more price-sensitive. The present four-company evidence is consistent with resilience, but it does not identify which bargaining outcome will dominate over several renewal cycles.

Investors should therefore look for the money at renewal: comparable customer-cohort spending, retained rights, price changes, seat or enterprise scope, and the cost of serving expanded usage. Buyers should ask the same questions from the opposite side of the contract.

Use a renewal bridge instead of an adoption multiplier

The illustrative tool starts with a customer cohort paying for 100 subscription units at 1,200 each. Suppose the next renewal covers 80 units at 1,300, plus a genuinely incremental, non-overlapping data-access fee of 20,000. Revenue rises from 120,000 to 124,000, or 3.33%, even though units fall by 20%.

The additional fee needed merely to preserve the earlier revenue would be 16,000. This example is hypothetical; none of the four companies disclosed this customer contract. Its purpose is to show why lower seat counts do not mechanically mean lower vendor revenue, and why higher unit prices do not prove stronger total economics.

The bridge must use the same customer cohort and rights perimeter. If the supposed additional fee already includes the units, adding both would double-count revenue. If data rights change, the model should display that change rather than call the entire difference pricing power.

For now, the evidence supports a measured conclusion. Financial-data providers are participating in agent distribution and reporting continuing recurring-business momentum. A durable AI pricing premium remains a hypothesis to test through renewal and contribution evidence, rather than a fact established by connecting the data to a model.

The evidence, side by side

Data Subscription Renewal Bridge

Hypothetical customer cohort, not a disclosed contract or forecast.

Old revenue must exceed zero. Same cohort, currency, period and rights perimeter. Additional fee must not include the unit charges already counted.

Scenarios to watch

Data owner captures more value

Conditional DN interpretation

Comparable renewal spending and contribution rise as agent distribution expands, after separating price, scope and volume.

Customer captures the benefit

Conditional DN interpretation

Usage expands under existing entitlements while vendor revenue remains stable. The product may be useful without a separate AI premium.

Interface pressure emerges

Conditional DN interpretation

Comparable customers reduce spending or negotiate broader rights at lower effective prices. More connector adoption does not prevent weaker economics.

Method and limitations

DN paired four financial reports with four official distribution announcements, covering LSEG, FactSet, MSCI and S&P Global. This is a four-provider case study, not a completed 20–30-provider panel. Financial periods differ: FactSet’s year/quarter ended 31 August 2026, while the other current results cover September. Growth metrics remain in their original categories and currency conventions; they are not averaged into one AI-growth rate. The MSCI bridge is 49.1/98.6=49.80%, showing the disclosed share of the total revenue increase attributable to asset-based fees. It does not attribute the remainder to AI. Cutoff: 13 November 2026 UTC.

The sources do not provide randomized deployment evidence, matched renewal cohorts or a complete price-volume decomposition. Provider-specific retention and run-rate definitions should not be ranked without reconciliation. Top-customer adoption is a selected population. Connector availability, a licence relationship and a successful financial quarter can coexist without proving causation. Corporate perimeter changes and future strategic transactions could alter comparability.

Primary sources and evidence register

Source dates and reporting periods differ. Several documents from one issuer are one source family, not independent confirmations. The evidence download contains the observations used in this article.

Frequently asked questions

Does using financial data through an AI assistant remove the data licence?

The examined LSEG, S&P Global and MSCI announcements retain licensed or entitlement-controlled access. Exact customer rights and charges depend on their agreements.

Do strong subscription results prove AI pricing power?

No. They support commercial resilience. A pricing-power claim needs comparable renewal, price, scope and volume evidence that separates AI from other business drivers.

Why separate MSCI’s asset-based fees?

Those fees have a different economic driver from recurring subscriptions. They supplied about half of the reported dollar revenue increase in the examined quarter, so total growth cannot be treated as an AI subscription measure.

What would change DN’s conclusion?

A repeatable customer-cohort bridge showing higher net spending and contribution after agent adoption would strengthen the pricing thesis. Comparable downgrades, reduced renewal spending or weaker contribution would challenge it.