BNY Mellon: Scaling an AI-enabled platform
Download case PDF (4 pages)Request teacher’s pack
Information cutoff: July 31, 2026
By the end of July 2026, BNY Mellon had made a substantial shift toward a platform operating model and had begun deploying artificial-intelligence tools across the firm. Its leaders described AI as a strategic opportunity for client service and efficiency. The same disclosures emphasized implementation, model, cyber, third-party and governance risks. The evidence available publicly did not show how much AI-specific value the company had captured.
For this classroom exercise, you are part of a hypothetical executive review panel asked to assess BNY’s FY2027-and-beyond pace and controls for AI-enabled workflows and unified platforms. The panel is fictional; BNY has not disclosed this specific vote, budget or decision. Your task is to present a resource-allocation recommendation and explain what evidence would change it.
From platform build to benefits
BNY presents itself as a global financial-services platform at the center of capital markets. Its businesses provide custody and administration, market and wealth services, investment management and related capabilities to institutional and wealth clients. In its 2025 Annual Report, management said more than 70% of its workforce was operating in platforms at year-end. The company described the transition as a multi-year effort to integrate services and simplify how it works with clients.
The July 2026 business update marked a new phase. BNY said the platform activation phase had been completed during the second quarter and that management was tracking toward approximately $500 million in incremental investment and $450 million in efficiency savings during 2026. The same presentation said AI integration had started to show benefits across the client lifecycle. The company did not say that the investment or savings figures were AI-specific, nor that the tracked savings had already been realized.
BNY also reported 160 enterprise AI solutions in production and 134 “digital employees,” which it defined as multi-agent AI solutions operating alongside people. All employees had access to Eliza, the company’s enterprise AI platform; BNY said Eliza could integrate models from multiple providers. The company reported 171,000 AI learning hours in 2025. These indicators describe the scale of deployment and training. Public materials did not report active use by workflow, error rates, validated productivity, project-level costs, or AI-attributable revenue and savings.
Results and alternatives for capital
BNY’s second-quarter 2026 results showed a strong business alongside continued expense growth. Revenue increased 13% year over year, including 11% growth in fee revenue and 20% growth in net interest income. Noninterest expense increased 7%. Management attributed much of the expense increase to revenue-related costs, investment and employee salary growth, partly offset by efficiency savings.
The company reported 606 basis points of positive operating leverage on a GAAP basis; adjusted operating leverage excluding notable items was 591 basis points, a non-GAAP measure. Pre-tax margin was 39.8% on a GAAP basis, and return on tangible common equity was 31.3% on a non-GAAP basis. These measures summarize the quarter; they do not identify the contribution of AI or the platform program.
BNY returned $1.5 billion to common shareholders during the quarter, including $1.1 billion of share repurchases. Its Standardized CET1 ratio was 11.0% at June 30. In July, the board approved a 19% increase in the quarterly common dividend, from $0.53 to $0.63 per share. A July presentation reported an 87% year-to-date total payout ratio. The figure is not a dividend payout ratio. The company’s capital returns make shareholder distributions a visible alternative use of resources, but public information does not establish how much capital is available for a particular project.
The results leave several explanations open. Revenue growth reflected more than platform execution: market values, client activity, net new business, reinvestment at higher yields and balance-sheet growth also mattered. Expense growth included investments and salaries, while efficiency savings partly offset it. The public data do not isolate AI’s effect on either side of the income statement.
Competitive and control context
State Street’s July 2026 earnings materials offered a peer reference. It reported 17% revenue growth, 5% GAAP expense growth (10% excluding notable items), a 10.8% CET1 ratio, and $631 million returned to common shareholders. Its CEO described an ongoing technology- and AI-enabled transformation. The company presented an approximately $1 billion annual run-rate transformation-benefit target by 2029, comprising about $750 million in productivity savings and $250 million in revenue. This was a management target for a broader transformation, not realized savings or an AI-only figure. State Street also announced a tokenized-fund servicing capability in Luxembourg with an initial client. Differences in business mix, scale, definitions and strategy make the comparison directional rather than a benchmark.
BNY’s Annual Report identified risks from implementing AI, failing to achieve expected productivity, biased or inaccurate outputs, cyber and system failures, oversight and accountability, third-party models and data, legal and regulatory changes, and the adequacy of governance and controls. Its July filing also warned that failure to adopt appropriate AI solutions, build capabilities throughout its platforms or realize expected benefits could weaken its competitive position and long-term performance. The disclosures describe risks and management’s view of the stakes; they do not report a specific AI incident or quantify the probability of those outcomes.
Academic research supplies mechanisms and task-specific evidence, but no return estimate for BNY. A model of general-purpose technology argues that process redesign, training and other intangible investment can precede measurable productivity gains; its U.S. historical analysis concerns computer-related intangibles, not an individual AI program (Brynjolfsson, Rock and Syverson, AEJ: Macroeconomics, 2021, pp. 333–372; NBER Working Paper No. 25148, working-paper mechanism pp. 3–18). In one staggered workplace rollout, access to a generative-AI assistant raised resolved support chats per hour by 0.301, reported as 15.2% relative to a 1.97 pretreatment mean; gains varied by worker experience and skill, and surveyed customer satisfaction did not significantly change (Brynjolfsson, Li and Raymond, QJE, 2025, Table II p. 907, Table III pp. 909–910, heterogeneity pp. 910–916). The support setting was one business-process software firm, not a diversified custody and asset-servicing platform.
Your decision
For the exercise, the review panel must give its recommendation for the pace and governance of incremental FY2027-and-beyond AI-enabled workflow and platform investment. Consider three broad paths:
- Accelerate: expand deployment and workflow integration to pursue client, capacity and productivity opportunities.
- Stage and gate: continue investment while requiring evidence on adoption, quality, client outcomes, savings and control readiness before scaling each workflow.
- Constrain incremental growth: limit additional discretionary investment and emphasize expense discipline and capital returns until stronger evidence is available.
These are classroom alternatives, not options disclosed by BNY. The sources do not identify which projects or expenses are mandatory, and no option should assume that safety, regulatory, cybersecurity or client obligations can be ignored.
Prepare a short recommendation that states what you would fund or defer, what measures would determine success, how you would weigh client and employee outcomes against expense and control risk, and what information you would request before authorizing wider deployment. You may propose stage gates, but any numerical thresholds or budgets must be labeled as your own assumptions; BNY has not published an AI-specific financial hurdle.
Discussion questions
- What does BNY’s public evidence establish about platform and AI deployment—and what remains unmeasured?
- What explains the difference between revenue growth and expense growth in Q2 2026? How much, if any, can be attributed to AI or the platform shift?
- What should count as a benefit: expense reduction, capacity, service quality, client retention, new revenue, risk reduction, or some combination? Who should validate each measure?
- How should BNY weigh a possible cost of slower execution against implementation, model, cyber and governance risks?
- What would make a staged approach credible? What evidence would justify expanding, redesigning or stopping a workflow?
- How useful is State Street’s transformation as a competitive signal, and where does the comparison break down?
Exhibits and source notes
Exhibit 1 — BNY platform and AI indicators. BNY 2025 Annual Report, CEO letter printed p. V and “AI at BNY” printed pp. XI–XIII (PDF pp. 6, 12–14); Q2 2026 Quarterly Update Presentation slide 3 (PDF p. 2). The $500 million investment and $450 million efficiency-savings figures are management’s 2026 tracking outlook for the broader platform/transformation agenda, not AI-specific spending or realized savings.
Exhibit 2 — Q2 performance and capital context. BNY Q2 2026 Form 10-Q, MD&A printed pp. 4–5 and capital section beginning p. 33; July 15 earnings release PDF pp. 1–2 and 9; Q2 update presentation slides 2 and 5. The 606 bps reported operating leverage is GAAP; 591 bps adjusted excluding notable items is non-GAAP. ROTCE and total payout ratio definitions should be preserved as reported.
Exhibit 3 — Risk, peer and research context. BNY 2025 Annual Report printed pp. 81–82 and 88; State Street Q2 2026 release pp. 1 and 4; State Street presentation slides 3, 15, 17 and footnote 35 on slide 28. State Street’s approximately $1 billion 2029 transformation-benefit figure is a forward-looking management target for work broader than AI. Academic locators: Brynjolfsson, Rock and Syverson, AEJ: Macroeconomics 13(1) (2021), pp. 333–372; Brynjolfsson, Li and Raymond, QJE 140(2) (2025), Table II p. 907, Table III pp. 909–910 and heterogeneous effects pp. 910–916. These studies do not estimate BNY’s return.
All company figures and statements above were available by July 31, 2026. Additional academic studies and instructor analysis are reserved for class discussion; students are not given a company-specific productivity coefficient or return estimate.