The Bank AI Disclosure Audit: How JPMorgan, Bank of America, Wells Fargo, Citigroup, Goldman Sachs, Morgan Stanley, U.S. Bancorp, PNC, Capital One, and Truist Disclose AI in Customer Service, Investment Banking, Trading, Compliance, and Coding — and Why JPMorgan's $2 Billion Annual AI Spend and 200,000-User LLM Suite Has Built a Citation Lead No Peer Has Matched
The everything-pr disclosure-quality audit of the ten largest U.S. banks by deposits, scored on six signals: named AI use cases, named partners, named accountable executives, peer-reviewed publication, regulatory filing detail, and public AI governance principles. Built fully in-house using only public-source research — SEC filings (10-K, 10-Q, 8-K, proxy statements), CEO shareholder letters, investor day presentations, corporate press releases, and trade press coverage from CNBC, Fortune, Business Insider, American Banker, Banking Dive, Bloomberg, and Wall Street Journal. Three banks lead the audit by structural margin. One Big Six bank trails its commercial scale most visibly — and the disclosure gap will compound through 2026 as AI-citation share continues consolidating around the top disclosers.
By April 2026, AI deployment in U.S. banking is no longer a strategic question; it is an operational one. Per Jamie Dimon's 2025 annual shareholder letter, published April 6, 2026: "The importance of AI is real, and while I hesitate to use the word transformational, it is. The pace of adoption will likely be far faster than prior technological transformations, like electricity or the internet. Those took decades to roll out, but this implementation looks likely to accelerate over the next few years."
Per JPMorgan's investor day disclosures filed with the SEC, the bank operates more than 100 generative AI solutions in production with 200,000+ employees using its internal LLM Suite. Per Bloomberg TV reporting from October 2025, Dimon said JPMorgan spends approximately $2 billion per year on AI and finds an equal amount of cost savings as a result; the bank forecasts $19.8 billion technology spending in 2026, up from $18 billion in 2025. Per Bank of America's April 2025 disclosure, Erica has surpassed 2.5 billion client interactions with 20 million active users, and over 90% of Bank of America's 213,000 employees use Erica for Employees with the virtual assistant having reduced calls into the IT service desk by more than 50%. Per Goldman Sachs's January 2025 launch, GS AI Assistant rolled out to 10,000 employees with full firm-wide deployment to 46,000+ employees by mid-2025.
The disclosure variation across the named ten banks is meaningful. Banks that disclose comprehensively are positioned for AI-citation share, recruiting, and analyst coverage advantages. Banks that disclose less are positioned with thinner records under regulatory scrutiny from the OCC, Federal Reserve, FDIC, and CFPB — all of which have published AI/algorithmic decision-making guidance applicable to bank operations.
This audit answers six questions bank communications teams should be running quarterly. Methodology and full scoring rubric at the bottom. Any reader can reproduce the score using only the public sources cited.
The everything-pr Bank AI Disclosure Audit — methodology
Six signals, 100-point composite scale.
Signal 1 (20 points): Named AI use cases publicly disclosed
Specific named AI tools, internal platforms, or customer-facing systems with disclosed scope, scale, and adoption metrics.
Signal 2 (15 points): Named AI partners with deal terms publicly disclosed
Specific partner companies, deal values, and program scope.
Signal 3 (15 points): Named accountable executives publicly identified
Chief Information Officer, Chief Data and Analytics Officer, Chief AI Officer, or named technology leadership.
Signal 4 (15 points): Peer-reviewed publication or research output
Bank-affiliated AI research published in NeurIPS, ICML, or other peer-reviewed venues.
Signal 5 (20 points): Regulatory disclosure detail
SEC 10-K AI disclosure depth, OCC supervisory guidance compliance, Federal Reserve disclosure, CFPB compliance documentation.
Signal 6 (15 points): Public AI governance principles document
Published responsible AI framework, model risk management documentation, or named human-in-the-loop policies.
Composite below 60 triggers Disclosure Risk tagging. Composite below 45 triggers Critical Disclosure Risk.
The scorecard
RankBankUse CasesPartnersExecutivesPeer-ReviewedRegulatoryGovernanceComposite1JPMorgan Chase201414131813922Bank of America191314111713873Morgan Stanley181414111612854Goldman Sachs181313111612835Capital One171212121511796Wells Fargo15111291411727Citigroup14111191411708U.S. Bancorp13101181310659PNC Financial Services1291071396010Truist Financial118106129
The deep audit
1. JPMorgan Chase — Composite 92
JPMorgan Chase leads the bank audit by structural margin, anchored by the most-disclosed AI program in U.S. financial services and named CEO commentary that has become an industry-wide reference document.
Named AI use cases (20/20). LLM Suite — debuted in 2024 per CNBC's August 2024 reporting, now reaching 200,000+ employees per JPMorgan's investor day disclosure. Per CNBC's September 2025 deep coverage, 250,000 JPMorgan employees have access to LLM Suite — "the entire workforce except for branch and call center staff." LLM Suite functions as a portal allowing users to tap external large language models from OpenAI, Anthropic, and others. Documented use cases: writing emails and reports, summarizing lengthy documents, problem-solving with Excel, generating ideas, marketing content for social media, mapping client travel itineraries, summarizing meetings, drafting investment banking pitch decks (per CNBC's demonstration: "creating a credible-looking PowerPoint deck in about 30 seconds" for a Nvidia banker pitch). Ask David — multi-agent AI system per Business Chief's reporting: "David standing for data, analytics, visualisation, insights and decision-making — which processes data to automate multi-step tasks in investment research." COIN (Contract Intelligence) — the firm's machine-learning contract analysis system, in production since 2017. Document Intelligence for SEC filing analysis and report generation. Coding assistance through generative AI tools deployed to JPMorgan's developer workforce. Per the Quartz coverage, JPMorgan President Daniel Pinto valued AI use cases at $1B-$1.5B in May 2024.
Named partners (14/15). OpenAI (LLM Suite initial deployment); Anthropic; Google (Gemini integration); Microsoft (cloud infrastructure); Nvidia (GPU compute partnership); Devin (CNBC reference). The partner disclosure is the deepest in the audit.
Named executives (14/15). Jamie Dimon — Chairman and CEO; named annual shareholder letter commentary on AI is industry-leading. Daniel Pinto — President and Chief Operating Officer (retired 2025), publicly quoted on AI valuation: "very, very impactful for the bank's 60,000 developers and 80,000 operations and call-center employees." Teresa Heitsenrether — Chief Data and Analytics Officer; quoted in CNBC, Business Insider, Yahoo Finance: "Ultimately, we'd like to be able to move pretty fluidly across models depending on the use cases. The plan is not to be beholden to any one model provider." Derek Waldron — Chief Analytics Officer (former McKinsey partner with PhD in computational physics); demonstrated LLM Suite to CNBC and articulated the firm's AI vision: "Every employee will have their own personalized AI assistant; every process is powered by AI agents, and every client experience has an AI concierge." Lori Beer — Chief Information Officer.
Peer-reviewed publication (13/15). JPMorgan AI Research, the bank's research group, publishes regularly in NeurIPS, ICML, and other peer-reviewed venues. Manuela Veloso — Head of AI Research, Herbert A. Simon University Professor (Carnegie Mellon University). The peer-reviewed track record is the deepest of any U.S. bank.
Regulatory disclosure (18/20). Industry-leading SEC 10-K AI disclosure depth; investor day AI presentations filed with SEC; named OCC, Federal Reserve, and CFPB compliance documentation; Senate-cooperative engagement on AI use disclosure.
Public governance (13/15). Published responsible AI principles; model risk management documentation aligned with SR 11-7; named human-in-the-loop policies. Per Heitsenrether: "Since our data is a key differentiator, we don't want it being used to train the model. We've implemented it in a way that we can leverage the model while still keeping our data protected." The structural framing is the industry-leading articulation.
2. Bank of America — Composite 87
Bank of America's AI disclosure is anchored by Erica — the most-cited single AI deployment in U.S. banking — with seven years of continuous disclosure cadence and quantified milestone reporting.
Named AI use cases (19/20). Erica — the AI virtual financial assistant launched 2018; per BofA's 2025 newsroom disclosure, Erica has surpassed 3 billion client interactions, with 50 million users since launch and 58 million interactions per month. The 2025 disclosure documents 1.7 billion proactive personalized insights delivered to clients. Per BofA's April 2025 disclosure, 2.5 billion interactions total at that time with 20 million active users. Erica for Employees — launched 2020; 90%+ of 213,000 employees use it; reduced IT service desk calls by 50%+. CashPro Chat — virtual service advisor for commercial clients; 65% of clients use it; Erica handles 40%+ of client interactions. ask MERRILL and ask PRIVATE BANK — Merrill and Private Bank tools leveraging Erica technology; 23 million interactions in 2024 (1 million increase over 2023). The Academy — AI-powered conversation simulators for employee onboarding; over 1 million simulations completed in 2024. BankerAssist — AI virtual assistant for business bankers leveraging Erica's underlying technology. Coding assistance — software developers using generative AI tool with 20%+ efficiency gains.
Named partners (13/15). Microsoft (cloud infrastructure); multiple LLM partners; Salesforce. BofA's published philosophy per CIONews coverage: "We don't want to be wedded to any given model" — Hari Gopalkrishnan. The disclosure depth is structural.
Named executives (14/15). Aditya Bhasin — Chief Technology and Information Officer; quoted in multiple disclosures: "AI is having a transformative effect on employee efficiency and operational excellence. Our use of AI at scale and around the world enables us to further enhance our capabilities, improve employee productivity and client service, and drive business growth." Hari Gopalkrishnan — CIO and head of Consumer, Business and Wealth Management Technology; quoted in CIODive coverage: "Erica has been learning from our clients for many years, enabling us to leverage AI today at scale, globally." Nikki Katz — Head of Digital. David Tyrie — Chief Digital Officer and Head of Global Marketing. Brian Moynihan — Chairman and CEO.
Peer-reviewed publication (11/15). BofA produces substantial AI research output; the peer-reviewed footprint is solid but trails JPMorgan AI Research's NeurIPS/ICML cadence.
Regulatory disclosure (17/20). Strong SEC 10-K AI disclosure; OCC and Federal Reserve compliance; published model risk management framework. The 213,000-employee Erica for Employees adoption metric is the most-quantified disclosure in U.S. banking.
Public governance (13/15). Published "human oversight, transparency, and accountability for all outcomes" framework. The annual AI investment disclosure ($4 billion in 2025 per CIONews coverage) is the most-quantified single-year AI spending disclosure in banking after JPMorgan.
3. Morgan Stanley — Composite 85
Morgan Stanley's AI disclosure is concentrated on the AI @ Morgan Stanley product family — Assistant, Debrief, AskResearchGPT — with 98% adoption among financial advisor teams and named OpenAI strategic partnership.
Named AI use cases (18/20). AI @ Morgan Stanley Assistant — launched September 2023, first Wall Street firm to deploy a bespoke GPT-4 solution per CNBC; 98% adoption among financial advisor teams per Morgan Stanley's June 2024 disclosure; document retrieval efficiency increased from 20% to 80%; access to "100,000 documents" library per OpenAI's case study (David Wu quote: "We went from being able to answer 7,000 questions to a place where we can now effectively answer any question from a corpus of 100,000 documents"). AI @ Morgan Stanley Debrief — launched June 2024 for ~16,000 financial advisors; uses Whisper and GPT-4 to summarize client meetings, draft follow-up emails, save notes to Salesforce; advisors save approximately 30 minutes of administrative work per meeting per Klover.ai analysis. AskResearchGPT — launched October 2024 for Investment Banking, Sales & Trading, and Research staff; provides access to 70,000+ proprietary research reports published annually; one-click email-draft workflow patented by Morgan Stanley. Eval framework — translation evals for multilingual clients, summarization evals with advisor-prompt-engineer grading.
Named partners (14/15). OpenAI (sole wealth management strategic partner since March 2023, expanded multiple times since). Salesforce (CRM integration for Debrief). The OpenAI exclusive wealth-management arrangement is the most-disclosed banking AI partnership of its kind.
Named executives (14/15). Andy Saperstein — Co-President; quoted in CNBC reporting: "Financial advisors will always be the center of Morgan Stanley wealth management's universe." Jeff McMillan — Head of Firmwide Artificial Intelligence (originally Head of Analytics, Data and Innovation); industry-leading public commentary: "This technology makes you as smart as the smartest person in the organization. Each client is different, and AI helps us cater to each client's unique needs." David Wu — Head of Firmwide AI Product & Architecture Strategy. Vince Lumia — Head of Morgan Stanley Wealth Management Client Segments. Ted Pick — CEO. James Gorman — former CEO and Executive Chairman. Daniel Simkowitz — Co-President.
Peer-reviewed publication (11/15). Morgan Stanley publishes substantial research; the AI-method-specific peer-reviewed publication track is moderate.
Regulatory disclosure (16/20). Strong SEC 10-K AI disclosure; FINRA cooperation; published OpenAI relationship structure with embedded "AI is subject to limitations" disclaimers. The disclosure of "$64 billion in net new assets in a single quarter" attributable in part to AI capacity creation per Klover.ai analysis is a quantified ROI disclosure rare in banking.
Public governance (12/15). Published OpenAI relationship structure with quality-assurance framework, zero data retention policy disclosure, and embedded compliance documentation.
4. Goldman Sachs — Composite 83
Goldman Sachs's AI disclosure is anchored by GS AI Assistant and CIO Marco Argenti's industry-leading public commentary.
Named AI use cases (18/20). GS AI Assistant — debuted internally 2024, full firm-wide deployment to 46,000+ employees announced June 2025 per Fortune coverage; built on the proprietary GS AI Platform (debuted 2024) with multi-LLM access including OpenAI, Gemini, Llama. Per CNBC's January 2025 reporting: the assistant handles drafting research notes, summarizing regulatory content, answering client queries, generating code, translating research documents into multiple languages. GS AI Platform — multi-modal foundation supporting all firm AI deployments. Coding agents — per Lucidate's July 2025 coverage, Goldman announced plans to deploy thousands of autonomous AI software engineers working alongside the bank's nearly 12,000 human developers, with 3-4x productivity gains projected per Argenti. Devin integration — Cognition Labs's AI software engineer deployed at Goldman scale.
Named partners (13/15). OpenAI; Google (Gemini); Meta (Llama); Cognition Labs (Devin); multiple AI infrastructure partners. Per Fortune's March 2025 coverage, Argenti's published philosophy: "I don't want to rely on just one vendor and is giving the firm the flexibility to use a model that may be better for coding, while a rival offering is stronger at reasoning."
Named executives (13/15). Marco Argenti — Chief Information Officer (former AWS VP of Technology); industry-leading public AI commentary including five-prediction Goldman Sachs research piece on 2025 AI evolution. Per CNBC: "The AI assistant becomes really like talking to another GS employee... actually reason more and become more like the way a Goldman employee would think." David Solomon — CEO; quoted publicly on AI: "enormous opportunities for productivity gains and also opportunities for efficiency." George Lee — Co-head of Goldman Sachs Global Institute (former co-CIO). John Waldron — President and Chief Operating Officer.
Peer-reviewed publication (11/15). Goldman publishes substantial research; the peer-reviewed AI publication track is moderate.
Regulatory disclosure (16/20). Strong SEC 10-K AI disclosure; Argenti's Goldman Sachs Insights publications constitute industry-leading public-facing thought leadership filed in regulator-watched venues.
Public governance (12/15). Published AI steering group and risk and control teams structure; named change management approach.
5. Capital One — Composite 79
Capital One's AI disclosure positions the company as the bank closest to a pure technology-company AI program — anchored by its long-running cloud migration, AWS partnership, and machine learning research arm.
Named AI use cases (17/20). Eno — virtual assistant launched 2017, predates most peer bank deployments. Capital One Machine Learning Research; fraud detection AI; credit decisioning AI; customer service AI; coding assistance AI. Per Evident's financial services AI index, Capital One ranked second only to JPMorgan in 2025 banking AI rankings.
Named partners (12/15). AWS (deep cloud-and-AI partnership); multiple AI vendor relationships; the cloud-native posture is publicly disclosed in 10-K.
Named executives (12/15). Richard Fairbank — Founder, Chairman, and CEO. Andrew Young — Chief Information Officer. Prem Natarajan — Chief Scientist and Head of Enterprise AI.
Peer-reviewed publication (12/15). Capital One AI Research publishes in NeurIPS and other peer-reviewed venues; the cadence is substantial relative to bank scale.
Regulatory disclosure (15/20). Strong SEC 10-K AI disclosure; OCC supervisory engagement.
Public governance (11/15). Published responsible AI framework; model risk management documentation.
6. Wells Fargo — Composite 72
Wells Fargo's AI disclosure is anchored by Fargo virtual assistant and the broader technology investment under CEO Charlie Scharf's modernization agenda.
Named AI use cases (15/20). Fargo — Wells Fargo's AI virtual assistant; per Wells Fargo disclosures, surpassed 245 million interactions in 2024. AI for fraud detection in commercial banking; AI-enhanced underwriting; coding assistance.
Named partners (11/15). Google Cloud (Fargo deployment); Microsoft; multiple AI vendor relationships.
Named executives (12/15). Charlie Scharf — CEO. Saul Van Beurden — CIO. Tracy Kerrins — Head of Generative AI for Consumer & Small Business Banking.
Peer-reviewed publication (9/15). Limited peer-reviewed AI publication.
Regulatory disclosure (14/20). SEC 10-K AI disclosure; ongoing OCC compliance disclosure given the consent order context.
Public governance (11/15). Published AI governance framework.
7. Citigroup — Composite 70
Citigroup's AI disclosure is concentrated on transformation initiatives under CEO Jane Fraser and CIO leadership.
Named AI use cases (14/20). Citi Stylus — generative AI for document processing. Citi Assist — internal employee assistant. AI for fraud detection, compliance, and client onboarding.
Named partners (11/15). Microsoft; multiple AI vendor relationships.
Named executives (11/15). Jane Fraser — CEO. Vis Raghavan — Head of Banking. Tim Ryan — Head of Technology and Business Enablement.
Peer-reviewed publication (9/15). Limited peer-reviewed AI publication.
Regulatory disclosure (14/20). SEC 10-K AI disclosure; ongoing regulatory transformation context.
Public governance (11/15). Published AI governance framework.
8-10. U.S. Bancorp (65), PNC (60), Truist (56)
These three regional super-banks operate substantial AI deployment but at materially thinner public-facing disclosure depth than the Big Six.
U.S. Bancorp — composite 65. CEO Andy Cecere; CIO Dominic Venturo. Disclosed AI deployment in business banking, commercial banking, and wealth management. The disclosure depth trails the Big Six.
PNC Financial Services — composite 60. CEO Bill Demchak; CIO Ganesh Krishnan. Sits exactly at the Disclosure Risk threshold.
Truist Financial — composite 56, triggers Disclosure Risk. CEO Bill Rogers; CIO Scott Case. Truist's post-merger technology integration has produced a thinner public-facing AI disclosure footprint than peer regional super-banks. The structural cause: Truist's BB&T-SunTrust integration consumed technology bandwidth that AI disclosure would have required.
Cross-bank patterns
Pattern 1: CEO commentary is the highest-leverage signal. Jamie Dimon's annual shareholder letter, Brian Moynihan's investor commentary, and Marco Argenti's published Goldman Sachs Insights pieces produce disproportionate trade-press lift. Banks without named CEO-level AI commentary face citation gaps no other disclosure can close.
Pattern 2: Named accountable AI executives drive citation depth. Teresa Heitsenrether (JPMorgan), Aditya Bhasin (BofA), Jeff McMillan (Morgan Stanley), Marco Argenti (Goldman) — these names appear in trade press at frequencies an order of magnitude higher than unnamed AI leaders at peer banks. The disclosure choice matters operationally.
Pattern 3: Quantified adoption metrics drive trade-press citation. "200,000 employees using LLM Suite" (JPMorgan), "98% of advisor teams using AI Assistant" (Morgan Stanley), "3 billion Erica interactions" (BofA), "46,000 employees with GS AI Assistant" (Goldman) — quantified disclosure produces direct trade-press lift. Banks that disclose without quantification get summarized; banks that quantify get cited verbatim.
Pattern 4: Investor day AI presentations are now mandatory. JPMorgan's $2B annual AI spend and 200,000-user LLM Suite disclosure both come from investor day presentations filed with the SEC. Banks that do not include AI-program detail in investor day presentations face analyst questions about disclosure depth.
Pattern 5: Multi-LLM strategy disclosure is the new industry-standard. JPMorgan's "not beholden to any one model provider" framing (Heitsenrether), Goldman's multi-LLM platform with OpenAI/Gemini/Llama, BofA's "We don't want to be wedded to any given model" (Gopalkrishnan), Morgan Stanley's expanded OpenAI relationship — all reflect the disclosure expectation that banks operate model-agnostic AI infrastructure rather than single-vendor lock-in.
Pattern 6: Coding-assistance productivity disclosure is the emerging metric. Goldman's 3-4x developer productivity projection, JPMorgan's coding assistance deployment, BofA's 20%+ developer efficiency gains. Banks that disclose coding-assistance metrics get cited; banks that do not face citation gaps as the trade press writes the developer-productivity story across 2026.
The everything-pr Bank AI Disclosure Standard
Five elements every bank should implement within 90 days.
Publish CEO-level AI commentary annually. Jamie Dimon's 2025 shareholder letter is the canonical model. Banks without CEO-level AI articulation face structural disclosure gaps no other communication channel can replace.
Disclose quantified AI adoption metrics quarterly. Employees with access; user-count by deployment; interaction-volume metrics; productivity-gain estimates. The JPMorgan investor day model is reproducible.
Name accountable AI executives publicly. Chief Data and Analytics Officer, Chief Information Officer, Head of Firmwide AI — with permanent web presence and quoted positions in trade press.
Publish multi-LLM strategy disclosure. Banks operating model-agnostic AI infrastructure should disclose the philosophy publicly. Single-vendor lock-in disclosure is now a structural disadvantage.
Publish a comprehensive responsible AI principles document. Aligned with SR 11-7, OCC supervisory guidance, and Federal Reserve model risk management framework.
The five elements are operational, not aspirational. Every bank in this audit can implement them within 90 days. The compounding citation advantage from disclosure compounds against banks that do not.
Methodology
The six signals are weighted as described above. Composite below 60 triggers Disclosure Risk; composite below 45 triggers Critical Disclosure Risk.
Data pulled from public sources: SEC filings (10-K, 10-Q, 8-K, proxy statements); CEO shareholder letters; investor day presentations filed with SEC; corporate press releases; OpenAI case studies; trade press from CNBC, Fortune, Business Insider, American Banker, Banking Dive, Bloomberg, Wall Street Journal, AIM Media House, eMarketer, Quartz, Healthcare-Brew, Klover.ai analyses, CIONews, Process Excellence Network, Lucidate, and ATM Marketplace. No paid databases used.
