Moody’s Corporation: Where Should AI’s Customer Interface Live?
Download case PDF (7 pages)Request teacher’s pack
Strategy case for MBA students and working professionals
Information cutoff: October 2, 2026, 4:00 p.m. ET
The allocation question
Moody’s combines a credit-ratings business with Moody’s Analytics (MA), which sells data, research, software and financial workflows. MA was investing in generative AI across its own applications and customer systems, while also making Moody’s information and analytical tools available through technology partners. By October 2026, the strategic choice was no longer whether to experiment with AI. It was how to allocate scarce product, engineering, data, sales and risk-control capacity among three routes:
- Build and sell more Moody’s-owned workflow suites.
- Embed Moody’s data and analytical capabilities inside customer and partner platforms.
- Stage a portfolio: own selected high-value workflows and distribute defined capabilities through multiple channels, scaling only as commercial and risk evidence develops.
Each route could make Moody’s information more useful. They differ in who controls the interface, customer relationship, implementation, usage data and associated costs. Public information documents launches, trials and management’s commercial claims, but does not disclose paid adoption, partner economics or AI-specific margins in a way that resolves the choice.
The task is to recommend how MA should allocate its next increments of investment—and what evidence should cause the company to change course.
A recurring analytics business with different growth signals
Moody’s Investors Service (MIS) earns revenue from ratings and related activities, including debt issuance that varies with market conditions. MA sells recurring data, analytics, software and workflow products. Its FY2025 external revenue was $3.599 billion, up 9% from 2024; adjusted operating income was $1.196 billion and adjusted operating margin was 33.1%. These adjusted figures are company-defined non-GAAP measures. [S1]
For the quarter ended June 30, 2026, MA reported $925 million in revenue, up 4% year over year. Organic constant-currency growth was 8%, a management-defined measure that excludes specified currency and portfolio effects. Recurring revenue was $915 million, or 99% of MA revenue, up 7% reported and 9% organic constant currency. MA’s annualized recurring revenue (ARR) was $3.661 billion, up 9%. ARR is management’s point-in-time estimate for recurring contracts; it is not recognized revenue, an AI-only measure, or a reported bookings backlog. [S2]
Exhibit 1. MA Q2 2026 revenue and ARR by line of business
| Business | Q2 revenue | Reported revenue growth | Organic constant-currency growth | ARR at June 30 | ARR growth |
|---|---|---|---|---|---|
| Decision Solutions | $423m | 2% | 12% | $1.621bn | 10% |
| — Banking | $119m | (14%) | 17% | — | — |
| — Insurance | $183m | 9% | 9% | — | — |
| — KYC | $121m | 13% | 11% | — | — |
| Research & Insights | $256m | 3% | 2% | $1.037bn | 6% |
| Data & Information | $246m | 9% | 8% | $1.003bn | 8% |
| Moody’s Analytics | $925m | 4% | 8% | $3.661bn | 9% |
Source: Moody’s Q2 2026 earnings release. Growth rates are year over year. Organic constant-currency growth is company-defined; portfolio changes and foreign exchange affect comparisons. Components may not sum due to rounding. The quarter is the latest reported financial period available by the case cutoff. [S2]
Reported and organic results give different pictures in some lines. Banking’s reported revenue declined while its organic constant-currency measure grew, reflecting factors such as portfolio and currency effects. MA’s results are not a direct measure of AI demand: Moody’s does not report AI-specific revenue, ARR, customer counts, usage or margins. Meanwhile, its portfolio had changed. It had sold Learning Solutions in 2025 and divested its Regulatory Solutions business in 2026; management described those businesses as more transactional or less connected to cross-selling with other banking products. [S1, S2]
Two delivery routes—and a staged alternative
Moody’s sells products through its own applications, including CreditLens and the Lending Suite, Moody’sView, and the Intelligent Risk Platform. It also offers data and analytical capabilities through APIs, MCP connections, customer systems and partner platforms. MCP servers provide a way for external AI systems to call Moody’s data and tools; the company describes them as complements to its reusable analytical “skills.”
The routes imply different operating models. An owned suite can place data, analysis, controls and task flow in one Moody’s product. It may give the company more direct control over packaging, customer interaction and feedback, while requiring customers to adopt another application or change existing processes. Embedded distribution may meet users in a familiar environment and reach additional buyer groups, while increasing dependence on platforms and raising questions about customer ownership, rights, usage visibility and the split of economics. These are strategic hypotheses; the public record does not establish which route earns better returns.
Exhibit 2. Product and distribution chronology
| Date | Public evidence | What it establishes |
|---|---|---|
| June 8, 2026 | In an investor Q&A, MA’s generative-AI leader described partnerships with AWS, Anthropic, Microsoft and OpenAI. She said partners could expand access to buyer groups while Moody’s sought to preserve its customer relationship and intellectual property. She described different token-cost arrangements for direct and partner environments, minimum/base and consumption pricing, and a premium for MCP over raw APIs. [S3] | Management’s intended packaging and value-capture logic. The transcript does not establish representative contract terms, realized margins, customer savings, or cost and usage cohorts. |
| June 17, 2026 | Moody’s announced an initial set of platform-agnostic AI skills—Earnings Call Summary, Peer Analysis, Public Information Book, Rating Pitch and Sector Analysis—launching on Microsoft 365 Copilot Cowork. Moody’s said skills encode analytical steps and MCP connects them to Moody’s data. [S4] | Product direction and announced availability beginning on Microsoft. No paid-user counts, bookings, quality outcomes or economics were reported. |
| July 22, 2026 | Q2 results were released. On the earnings call, management said Moody’s data was available in Amazon Quick, more than 20 Microsoft engagements and initial trials were underway, and more than 100 MCP and Smart API connections were being used or trialed. [S2, S5] | A mix of platform availability, engagements, trials and connections. The 100-plus figure combines use and trials and is not a count of unique paying customers. These announcements do not change the quarter-end financial data. |
| August 25, 2026 | Moody’s announced that its Credit MCP server made connected intelligence available within Google Cloud’s Gemini Enterprise for Financial Services and identified itself as a launch partner. [S6] | An announced integration and availability claim. It does not disclose paid adoption, unique customers, usage, bookings, revenue share or economics. |
The sequence was: Moody’s announced platform-agnostic skills on June 17, with an initial Microsoft launch; on July 22, Q2 results were released and management described Amazon availability and Microsoft engagements and initial trials; on August 25, Moody’s announced the Gemini Enterprise integration. The releases point to multiple distribution surfaces. They do not show that skills are adopted across platforms, that Gemini access is paid, or that customers use the same Moody’s capability in all channels.
What the adoption claims say—and leave open
At the February 2026 discussion of Q4 2025 results, Moody’s chief executive said roughly two thirds of eligible 2025 renewals converted to its AI-enabled Lending Suite, at an average uplift of about 67%. Management also said customers buying or upgrading into at least one standalone GenAI or agentic solution were retained at 97% and growing at roughly twice the rate of the rest of the MA customer base. These are management-reported figures. The remarks do not define the eligible renewal base, cohort size, uplift, measurement period, comparison group or realized revenue contribution. They cannot establish a company-wide adoption rate or a causal price or retention effect. [S7]
Management also described selected examples, including a tier-one U.S. bank using agents to generate portions of credit memos. In Q2 it discussed the platform engagements and trials listed in Exhibit 2. The public materials do not provide customer-level usage, independent productivity measurements, error or review rates, paid conversion by channel, contract economics or AI-specific costs. [S2, S5, S7]
The distinction matters for investment. An announcement can establish that a product or connection is available. An engagement or trial suggests evaluation. Neither establishes recurring paid use, willingness to pay, renewal impact or positive contribution after data, compute, integration, support and partner costs. The company’s reported MA recurring revenue and ARR provide a business baseline, not proof of AI’s contribution.
Competition and control of the interface
Competitors are pursuing similar combinations of specialized financial information and AI access. S&P Global describes AI assistants in Capital IQ Pro and LLM-ready APIs for external models. LSEG has expanded Microsoft Copilot integrations and announced a research agent in Workspace and Teams. These company materials establish product direction, not comparative adoption, product performance or profitability. [S8, S9]
Moody’s describes proprietary data, linkages, analytical frameworks and auditability as sources of differentiation. Its filings also identify risks associated with inaccurate or insufficiently explainable AI outputs, data and intellectual-property rights, third-party platforms, cybersecurity and changing regulation. Management’s claim that particular data cannot be synthesized from public sources is a competitive argument, not an independently tested finding. The company also licenses third-party data, software and AI tools. [S1, S5]
The platform choice therefore concerns both access and control. If users prefer a familiar external workspace, embedded Moody’s data might make the core information more useful and widen reach. But if the platform owns the interaction and usage signals, Moody’s may have less visibility into customer needs or less ability to capture the value. An owned workflow might preserve more product control and support a complete task, but adoption friction and implementation requirements could limit reach. A portfolio approach could preserve options while adding coordination costs and product complexity.
Evidence from other settings
In a field study at one business-process software company, Brynjolfsson, Li and Raymond examined staggered access to a generative-AI assistant among 5,172 customer-support agents. The authors report an average 15% increase in issues resolved per hour. Effects varied by worker and task: less experienced agents benefited more, while the most experienced had smaller speed gains and some quality measures weakened. This setting was customer support, not regulated credit analysis; it does not estimate Moody’s customer ROI, willingness to pay, renewal effect or product performance. [A1]
Brynjolfsson and Hitt studied computerization in a panel of 527 large U.S. firms from 1987–1994. Their estimated output contribution over three-to-seven-year periods was two to five times the one-year contribution, which they interpret as consistent with complementary organizational investments taking time. The study predates cloud software and modern AI, is observational, and does not estimate returns for Moody’s or favor owned versus embedded distribution. [A2]
Together, the studies suggest questions to test: which tasks benefit, for which users, with what quality controls, and what data, integration, training and workflow changes are required? Neither paper supplies a productivity rate or return assumption for Moody’s.
The decision
Moody’s leaders must choose where the next increments of MA investment go. A predominantly owned-workflow strategy could concentrate resources on complete Moody’s applications for high-value tasks such as lending and credit research. A distribution-led strategy could prioritize MCP, APIs and integrations that place Moody’s information inside platforms customers already use. A staged portfolio could designate a few workflows for deeper ownership, expose selected capabilities through partners and set evidence gates before expanding either route.
Each path has unresolved questions. Which customers and workflows have a sufficiently costly problem? What implementation and governance burden accompanies adoption? Who pays, who bears token and integration costs, and who owns the customer relationship? How should Moody’s distinguish usage from trials and recurring paid adoption? What measures would reveal customer value while guarding against quality or control failures?
Assignment
You are advising Moody’s Analytics on its next phase of AI investment. Recommend whether it should prioritize owned workflow suites, embedded customer and partner distribution, or a staged portfolio. Explain:
- Which use cases and customer segments should receive investment first, and why?
- What evidence would you require before scaling a launch, trial or integration?
- How would you measure paid adoption, renewal, customer value, product quality and channel economics?
- How should data rights, auditability, human review, customer ownership and third-party dependence affect the design?
- What are the key unknowns, and what findings would change your recommendation?
Do not treat management’s renewal conversion or uplift figures as verified causal results. Separate product availability, trials, management claims and paid commercial outcomes. Do not assume that skills adoption is platform agnostic in practice or that the Gemini integration has paid users.
Sources
- [S1] Moody’s Corporation, 2025 Form 10-K, filed February 18, 2026.
- [S2] Moody’s Corporation, Q2 2026 earnings release, July 22, 2026.
- [S3] Moody’s, June 8, 2026 Generative AI strategy Q&A transcript, third-party transcript; see also the official event listing.
- [S4] Moody’s, Decision-grade AI skills announcement, June 17, 2026.
- [S5] Moody’s Q2 2026 earnings call, third-party transcript, July 22, 2026.
- [S6] Moody’s, Gemini Enterprise/Credit MCP announcement, August 25, 2026.
- [S7] Moody’s, Q4 2025 prepared remarks (PDF), February 18, 2026. The PDF filename, directory, date and content identify Q4 2025; its rendered first-page title says “3Q24,” an apparent labeling defect.
- [S8] S&P Global, 2025 Annual Report AI product showcase.
- [S9] LSEG, Microsoft partnership, March 26, 2026; Workspace Deep Research agent, June 2026.
- [A1] Erik Brynjolfsson, Danielle Li and Lindsey R. Raymond, “Generative AI at Work”, Quarterly Journal of Economics 140(2), 2025, pp. 889–942.
- [A2] Erik Brynjolfsson and Lorin M. Hitt, “Computing Productivity: Firm-Level Evidence”, Review of Economics and Statistics 85(4), 2003, pp. 793–808.